{ "pack_id": "pack_e4072ef8b88d", "schema": "regbench.exam-pack/1.0", "supplement_version": "NAIC AI Risk Evaluation Supplement v6.0 (exposure draft, 16 Oct 2026); v7.0 pending adoption vote, Fall National Meeting", "tenant_id": "bluebonnet", "company": "Bluebonnet Mutual Insurance Company (fictional)", "naic_cocode": "DEMO-10101", "state": null, "period": { "from": "2026-07-02T00:00:00+00:00", "to": "2026-10-31T00:00:00+00:00" }, "created_at": "2026-10-30T11:45:10.439990+00:00", "engine": { "engine_version": "regbench-engine/0.4.0", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "prototype_version": "0.4.0" }, "confidentiality": "Produced for the requesting state's examination or inquiry; to be held confidential under that state's authority. The carrier may assert trade-secret protection over model documentation it attaches.", "exhibit_a": { "title": "Exhibit A: Part One - Quantify Regulated Entity's Use of AI Systems", "header": { "company": "Bluebonnet Mutual Insurance Company (fictional)", "naic_cocode": "DEMO-10101", "contact": "AI Governance Office, compliance@bluebonnet.example", "lines_of_business": [ "auto", "home" ], "materiality_threshold": "Models touching more than 2% of annual written premium or any direct consumer outcome; rationale: FCEH financial-position test plus consumer-impact floor.", "months_window": 12, "date_completed": "2026-10-30" }, "areas": [ { "operation_area": "Marketing", "systems_in_use": 1, "implemented_in_past_months": 1, "use_cases": [ { "system_id": "agent-assist", "use_case": "Generative AI assistant drafting service responses for human agents; no decision authority. Observed 30 decisions in the period, 0 adverse to the consumer.", "direct_consumer_impact": false, "material_financial_impact": "immaterial" } ], "generative_ai": 1, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Premium Quotes & Discounts", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Underwriting/Eligibility", "systems_in_use": 1, "implemented_in_past_months": 0, "use_cases": [ { "system_id": "uw-triage", "use_case": "Scores new auto and home applications for eligibility and refers borderline risks to an underwriter. Observed 150 decisions in the period, 26 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "material" } ], "generative_ai": 0, "agentic_ai": 0, "glms": 1, "other_ai_ml": 0 }, { "operation_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "systems_in_use": 2, "implemented_in_past_months": 0, "use_cases": [ { "system_id": "price-opt", "use_case": "Recommends renewal premium within filed rating plan bounds using telematics, credit-based insurance score and territory. Observed 168 decisions in the period, 45 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "material" }, { "system_id": "uw-triage", "use_case": "Scores new auto and home applications for eligibility and refers borderline risks to an underwriter. Observed 150 decisions in the period, 26 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "material" } ], "generative_ai": 0, "agentic_ai": 0, "glms": 1, "other_ai_ml": 1 }, { "operation_area": "Claims/Adjudication", "systems_in_use": 2, "implemented_in_past_months": 1, "use_cases": [ { "system_id": "claims-fastpath", "use_case": "Observed in the decision stream only (not in the declared inventory): claims-fastpath. Observed 109 decisions in the period, 10 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "unknown" }, { "system_id": "fnol-fraud", "use_case": "Vendor model scoring first notice of loss for fraud indicators; high scores route to SIU and can hold payment. Observed 120 decisions in the period, 6 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "moderate" } ], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 2 }, { "operation_area": "Customer Service", "systems_in_use": 1, "implemented_in_past_months": 1, "use_cases": [ { "system_id": "agent-assist", "use_case": "Generative AI assistant drafting service responses for human agents; no decision authority. Observed 30 decisions in the period, 0 adverse to the consumer.", "direct_consumer_impact": false, "material_financial_impact": "immaterial" } ], "generative_ai": 1, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Utilization Management/Utilization Review/Prior Authorization/Level of Care Determination", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Fraud/Waste & Abuse", "systems_in_use": 1, "implemented_in_past_months": 0, "use_cases": [ { "system_id": "fnol-fraud", "use_case": "Vendor model scoring first notice of loss for fraud indicators; high scores route to SIU and can hold payment. Observed 120 decisions in the period, 6 adverse to the consumer.", "direct_consumer_impact": true, "material_financial_impact": "moderate" } ], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 1 }, { "operation_area": "Investment/Capital Management", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Reserves/Valuations", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Catastrophe Triage", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Reinsurance", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 }, { "operation_area": "Other Insurance Practices", "systems_in_use": 0, "implemented_in_past_months": 0, "use_cases": [], "generative_ai": 0, "agentic_ai": 0, "glms": 0, "other_ai_ml": 0 } ], "inventory": [ { "name": "Underwriting Triage (uw-triage)", "description": "Scores new auto and home applications for eligibility and refers borderline risks to an underwriter.", "program_area": "Underwriting/Eligibility", "inherent_risk": "high", "direct_consumer_impact": true, "financial_impact": "material", "declared_version": "2.4", "observed_versions": [ "2.4" ], "observed_decisions": 150, "observed_adverse": 26, "first_seen": "2026-07-06", "last_seen": "2026-10-27" }, { "name": "Renewal Price Optimizer (price-opt)", "description": "Recommends renewal premium within filed rating plan bounds using telematics, credit-based insurance score and territory.", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "inherent_risk": "high", "direct_consumer_impact": true, "financial_impact": "material", "declared_version": "7.1", "observed_versions": [ "7.2" ], "observed_decisions": 168, "observed_adverse": 45, "first_seen": "2026-07-02", "last_seen": "2026-10-27" }, { "name": "Sentinel FNOL Fraud Score (fnol-fraud)", "description": "Vendor model scoring first notice of loss for fraud indicators; high scores route to SIU and can hold payment.", "program_area": "Fraud/Waste & Abuse", "inherent_risk": "high", "direct_consumer_impact": true, "financial_impact": "moderate", "declared_version": "3.1", "observed_versions": [ "3.1" ], "observed_decisions": 120, "observed_adverse": 6, "first_seen": "2026-07-02", "last_seen": "2026-10-22" }, { "name": "Service Desk Assistant (agent-assist)", "description": "Generative AI assistant drafting service responses for human agents; no decision authority.", "program_area": "Customer Service", "inherent_risk": "low", "direct_consumer_impact": false, "financial_impact": "immaterial", "declared_version": "1.2", "observed_versions": [ "1.2" ], "observed_decisions": 30, "observed_adverse": 0, "first_seen": "2026-07-08", "last_seen": "2026-10-27" } ], "reconciliation": { "observed_not_declared": [ { "system_id": "claims-fastpath", "versions": [ "1.0" ], "vendor": "internal", "high_risk_class": "claims", "decisions": 109, "adverse": 10, "first_seen": "2026-07-02" } ], "declared_not_observed": [], "version_drift": [ { "system_id": "price-opt", "declared_version": "7.1", "observed_versions": [ "7.2" ] } ] }, "note": "Counts come from two sources: the carrier's declared inventory and RegBench's observed decision stream for the period. The reconciliation shows where they disagree." }, "exhibit_b": { "title": "Exhibit B: (Checklist) AIS Program", "q1_written_program": { "adopted": "2025-03-18", "approved_by": "Board Risk Committee", "owner": "Chief Compliance Officer", "review_frequency": "annual", "scope": "all AI systems that make or support regulated insurance decisions" }, "q2_board_role": { "involved": true, "role": "Board Risk Committee approved the program; receives RegBench operating-effectiveness reporting", "accountability_training": "carrier to supply", "materiality_and_risk_assessment": "Models touching more than 2% of annual written premium or any direct consumer outcome; rationale: FCEH financial-position test plus consumer-impact floor." }, "q3_checklist": [ { "ref": "3a", "item": "Considers AI System risks within its Enterprise Risk Management (ERM)", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3b", "item": "Considers AI System risks within the Own Risk and Solvency Assessment Report (ORSA), as applicable", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3c", "item": "Assesses, mitigates, and evaluates inherent risks related to the use of AI System including the possibility of the company engaging in unfair trade practices", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Exhibit C item 13 and the case files: every decision assessed against unfair-trade-practice and unfair-claims-settlement provisions in force on the decision date." }, { "ref": "3d", "item": "Determined 'materiality' and what threshold was used for assessing AI Systems risk and determined the company's risk assessment processes for its use of AI Systems", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3e", "item": "Quantifies AI System risk levels", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3f", "item": "Considers AI System risk impact on financial reporting", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3g", "item": "Ensures AI Systems are compliant with applicable state and federal laws and regulations, including non-discrimination protections", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Case files: cited findings by provision, layer and severity; rules and corpus versions stamped on each assessment." }, { "ref": "3h", "item": "Evaluates the risk of Adverse Consumer Impact", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Operating-effectiveness statistics: adverse decisions assessed, findings on adverse decisions, review outcomes." }, { "ref": "3i", "item": "Considers data privacy and protection of consumer data used in AI Systems", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3j", "item": "Considers consumer complaints resulting from AI Systems and whether they are identified, tracked, and addressed", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3k", "item": "Promotes consumer awareness of the use of AI Systems through disclosures, policies, and procedures for consumer notification, as appropriate", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Case files: notice status (sent, reasons stated, specific reasons, external-data disclosure) recorded per adverse decision; notice findings." }, { "ref": "3l", "item": "Trains employees about AI System use, ensures human oversight is appropriate, and defines prohibited practices (if any)", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Human-in-the-loop rates by system for direct-consumer-impact decisions; unreviewed adverse outcomes routed to review." }, { "ref": "3m", "item": "Provides standards and guidance for procuring and engaging AI System vendors", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Exhibit A reconciliation: vendor systems observed in the decision stream against the declared inventory." }, { "ref": "3n", "item": "Governs, monitors, tests and ensures transparency of AI Systems developed by vendors", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Third-party system findings (due diligence, documentation, testing references) per decision; Exhibit C for vendor models." }, { "ref": "3o", "item": "Considers AI System risks within the software development lifecycle (SDLC) or within the broader AIS Program", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Not evidenced by RegBench; carrier AIS Program reference required." }, { "ref": "3p", "item": "Evaluates whether AI Systems are suitable for their intended use and should continue to be used as designed", "document_name": "Bluebonnet Mutual Insurance Company (fictional) AIS Program", "page": "carrier to supply", "regbench_evidence": "Exhibit C item 10/10a: continuous assessment coverage and portfolio outcome tests per model." } ], "subject_matter_experts": [ "AI Governance Office, compliance@bluebonnet.example", "RegBench corpus desk (for assessment methodology questions)" ], "operating_effectiveness": { "period": { "from": "2026-07-02T00:00:00+00:00", "to": "2026-10-31T00:00:00+00:00" }, "assessments": 577, "dispositions": { "PASS": 464, "HOLD": 20, "FLAG": 93 }, "disposition_pct": { "PASS": 80.4, "HOLD": 3.5, "FLAG": 16.1 }, "findings_by_severity": { "high": 76, "medium": 157, "low": 27, "critical": 4 }, "findings_by_layer": { "L0": { "label": "Federal law", "count": 18 }, "L2": { "label": "State statute", "count": 29 }, "L3": { "label": "State regulation", "count": 26 }, "L4": { "label": "Bulletin / circular letter", "count": 60 }, "L5": { "label": "Enforcement precedent", "count": 5 }, "L6": { "label": "Doctrine (interpretive position)", "count": 126 } }, "top_rules": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "count": 70, "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "layer": "L6", "judgment_basis": "doctrine", "advisory": true }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "count": 56, "title": "Adverse outcome from a high-risk AI system with no human review", "layer": "L6", "judgment_basis": "doctrine", "advisory": true }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "count": 47, "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "layer": "L4", "judgment_basis": "guidance", "advisory": false }, { "rule_id": "co.reg-10-1-1.governance-framework-amended", "count": 26, "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "layer": "L3", "judgment_basis": "black_letter", "advisory": false }, { "rule_id": "tx.559.credit-adverse-effect-notice", "count": 14, "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false }, { "rule_id": "us.fcra.permissible-purpose-undocumented", "count": 14, "title": "Consumer report used with no data source reference recorded", "layer": "L0", "judgment_basis": "black_letter", "advisory": false }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "count": 13, "title": "Third-party data used with no source recorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false }, { "rule_id": "tx.542.algorithmic-denial-without-human-review", "count": 9, "title": "Algorithmic denial, reduction or fraud hold with no human review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false }, { "rule_id": "tx.542.rejection-without-reasons", "count": 6, "title": "Claim rejected or reduced without stated reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "count": 5, "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "layer": "L5", "judgment_basis": "precedent", "advisory": true }, { "rule_id": "us.fcra.adverse-action-notice", "count": 4, "title": "Adverse action based on a consumer report with no FCRA notice recorded", "layer": "L0", "judgment_basis": "black_letter", "advisory": false } ], "adverse_decisions": 87, "adverse_with_findings": 82, "direct_consumer_impact_decisions": 547, "human_in_the_loop": [ { "system_id": "claims-fastpath", "decisions": 109, "human_reviewed_pct": 3.7, "adverse": 10, "automation": "unknown" }, { "system_id": "fnol-fraud", "decisions": 120, "human_reviewed_pct": 1.7, "adverse": 6, "automation": "automate" }, { "system_id": "price-opt", "decisions": 168, "human_reviewed_pct": 19.0, "adverse": 45, "automation": "automate" }, { "system_id": "uw-triage", "decisions": 150, "human_reviewed_pct": 38.7, "adverse": 26, "automation": "augment" } ], "review_queue": { "opened": 113, "resolved": 102, "open": 11, "median_hours_to_resolve": 17.9, "outcomes": { "remediated": 57, "overridden": 12, "confirmed-no-change": 26, "escalated": 7 } }, "rules_and_corpus_versions_in_force": [ { "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521" } ], "ledger_integrity": { "ok": true, "entries": 690, "first_bad_seq": null, "head": "ecb444b6d2530cf62784043ce02a0e2591a68aa8108e2e685a473bef736e16d7", "reason": null } } }, "exhibit_c": { "title": "Exhibit C: High-Risk AI Models", "prepared_by": "RegBench (evidence assembly); carrier compliance function (responses)", "reviewed_by": "carrier to supply", "models": [ { "system_id": "claims-fastpath", "items": { "1_name_version": "claims-fastpath \u2014 versions observed: 1.0 (not in declared inventory)", "2_use_cases": "Observed only. Stages: {'claims': 109}; decisions in period: 109; adverse: 10; states: {'TX': 82, 'CO': 27}.", "3_model_type": "unknown (not declared)", "3a_agentic": { "agentic": false, "objectives": null, "authorizations_without_human_approval": null }, "4_implementation_date": "2026-07-02", "5_development": { "vendor": "internal", "vendor_name": null }, "6_purchased_modifiable": { "purchased": null, "developed_for_company": null, "modifiable": null }, "7_model_risks": { "description": "Potential adverse consumer impact evidenced by findings on the model's decisions in the period.", "findings": [ { "rule_id": "doctrine.unreviewed-adverse-high-risk", "title": "Adverse outcome from a high-risk AI system with no human review", "severity": "medium", "layer": "L6", "count": 6 }, { "rule_id": "tx.542.algorithmic-denial-without-human-review", "title": "Algorithmic denial, reduction or fraud hold with no human review", "severity": "high", "layer": "L2", "count": 5 }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "severity": "high", "layer": "L4", "count": 5 }, { "rule_id": "tx.542.rejection-without-reasons", "title": "Claim rejected or reduced without stated reasons", "severity": "high", "layer": "L2", "count": 4 } ], "adverse_decisions": 10, "dispositions": { "PASS": 102, "FLAG": 3, "HOLD": 4 } }, "8_limitations": { "carrier_documentation": "no model documentation reference recorded on the decision stream", "regbench_note": "RegBench assesses recorded facts about each decision; it does not inspect model internals or detect inputs outside the training domain." }, "9_automate_augment_support": { "declared": null, "observed_human_review_pct": 3.7 }, "10_testing_validation": { "testing_report_ref_on_decisions": [ "none recorded" ], "regbench_continuous_assessment": "109 of 109 decisions assessed at decision time against the rules in force on each decision date." }, "10a_proxy_evaluation": [ { "system_id": "claims-fastpath", "model_version": "1.0", "tenant_id": "bluebonnet", "state": null, "stage_mix": { "claims": 109 }, "reference_group": "A", "groups": [ { "group": "A", "decisions": 64, "adverse": 6, "adverse_rate": 0.0938 } ], "suppressed_groups": [ "D", "B", "C" ], "flags": [], "method": "Adverse-outcome rate by carrier-estimated demographic proxy group; reference group = largest group; adverse impact ratio = group rate / reference rate (flag below 0.80); marginal difference in percentage points (flag at 5.0 or more). Cells under 30 decisions are suppressed. Illustrative thresholds.", "computed_where": "carrier environment (proxy labels never leave it); only these aggregates are released" } ], "11_last_test_date": { "carrier": "see testing report reference", "regbench_last_assessment": "2026-10-27T00:30:39+00:00" }, "12_financial_statement_effect": "unknown", "13_legal_compliance_review": { "rules_applied": [ "doctrine.unreviewed-adverse-high-risk", "tx.542.algorithmic-denial-without-human-review", "tx.542.rejection-without-reasons", "tx.tdi.consequential-decision-without-human-review" ], "provisions_cited": [ "naic-ai-bulletin-s3-1-ais-program", "tx-ins-542-003", "tx-ins-542-055", "tx-ins-542-056", "tx-tdi-b-0003-26-governance", "tx-tdi-b-0003-26-human-review" ], "method": "Deterministic rule evaluation per decision; citations resolve against the corpus snapshot in force on the decision date or the assessment fails closed." }, "14_regulatory_actions": "None recorded in RegBench. Carrier to confirm any informal agreements, administrative complaints, consent orders or investigations." }, "case_files": [ { "txn_id": "BB-CLM-0543", "assessment_id": "ra_8f3797470bcc4838", "occurred_at": "2026-07-02T05:37:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f2be37931a" }, "output": { "decision": "pay", "score": 0.085, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-02, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.918740+00:00", "fingerprint": "f7fbabc9563753a4ad876dff" }, "review": null, "ledger": { "seq": 543, "hash": "3196120d4e8302d8cfebd958599363750573ffd6acadfeb39ce6bdc8b5bc4222", "prev_hash": "34dd9eed1d8138528f6769f5952e4d8acdd1d616db5f9aeb23cf665cf348713b", "recorded_at": "2026-07-02T05:41:13+00:00" } }, { "txn_id": "BB-CLM-0458", "assessment_id": "ra_da2d0f280f0541cf", "occurred_at": "2026-07-04T19:04:46+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_249b513093" }, "output": { "decision": "pay", "score": 0.135, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-04, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.665470+00:00", "fingerprint": "5e0575531f081cf158833480" }, "review": null, "ledger": { "seq": 458, "hash": "c5967ea4393843e0a127b5f1257fb8f7918f1bdc01d5a346ca8ed5a807e27aa6", "prev_hash": "38a9bac91f134dd4e9ba841650b6766e8ab488835c102f9f47fb437a74a2a516", "recorded_at": "2026-07-04T19:05:16+00:00" } }, { "txn_id": "BB-CLM-0492", "assessment_id": "ra_0d9f7685545246c4", "occurred_at": "2026-07-07T00:26:18+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_da2f660a7f" }, "output": { "decision": "pay", "score": 0.194, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-07, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.757754+00:00", "fingerprint": "9209f8cdbcaf11804160a43d" }, "review": null, "ledger": { "seq": 492, "hash": "f212b8522fe5ee5626bebcec59f892ecf1cc5d85535756dc5f9644c33ed465af", "prev_hash": "81817a4fe10dd8b693a12e9baa64e2242cc91c68150bd55ddff79baab849ab92", "recorded_at": "2026-07-07T00:29:39+00:00" } }, { "txn_id": "BB-CLM-0461", "assessment_id": "ra_487060563ade4f35", "occurred_at": "2026-07-07T19:17:13+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_9cbebd1cd6" }, "output": { "decision": "pay", "score": 0.149, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-07, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.677527+00:00", "fingerprint": "667cf37c447966adda7fb64f" }, "review": null, "ledger": { "seq": 461, "hash": "cb5e1e9b0868b751d652c654fc2073cd8570a32154f28f930d83e725ff741bf4", "prev_hash": "bee68ce7fb0d39a705708e2c227464f63c07d610dd2ed5ec722347ca19d200f9", "recorded_at": "2026-07-07T19:18:14+00:00" } }, { "txn_id": "BB-CLM-0481", "assessment_id": "ra_18d66a7e8b2a44ab", "occurred_at": "2026-07-08T01:07:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_22ceacedcc" }, "output": { "decision": "pay", "score": 0.127, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-08, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.720155+00:00", "fingerprint": "3faebbc6fe61447983495aba" }, "review": null, "ledger": { "seq": 481, "hash": "bc8ed3d37d2ecc753e0f9632dbf48fb6883bdfd288edcaaeea04e3a6c9d0aaa7", "prev_hash": "527dcf2f51f9dd2151736c7ab006f18a454f1fa9b3f4be8429a9912a99c0dc56", "recorded_at": "2026-07-08T01:11:15+00:00" } }, { "txn_id": "BB-CLM-0500", "assessment_id": "ra_f6b075d2ab0f4db7", "occurred_at": "2026-07-08T11:19:05+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a26a0b8767" }, "output": { "decision": "pay", "score": 0.017, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-08, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-07-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.773171+00:00", "fingerprint": "71cdbf464fa131a34273aa31" }, "review": null, "ledger": { "seq": 500, "hash": "c6490c95c9c5d8cafeb48389895eb8f0bae28584aad802f8ac2d9c43e9869c16", "prev_hash": "cc89974bc5a7f8b099b974bed1cb3171ba58883dfd50809515bc2d535baf215e", "recorded_at": "2026-07-08T11:21:18+00:00" } }, { "txn_id": "BB-CLM-0545", "assessment_id": "ra_a5596afb64694d3d", "occurred_at": "2026-07-10T06:41:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3dfe26356b" }, "output": { "decision": "pay", "score": 0.08, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.923699+00:00", "fingerprint": "e48bca3f9d216e8a6d92e914" }, "review": null, "ledger": { "seq": 545, "hash": "5a7ee22fca77dc0a308bbc99f069d62ecb130c41056bb4bde412b2a4c41b4d62", "prev_hash": "90c767c1820cd246fc4c7a04b4cd1e7df7806f1d604b5bffdf1b2c56c82f959b", "recorded_at": "2026-07-10T06:42:57+00:00" } }, { "txn_id": "BB-CLM-0515", "assessment_id": "ra_91ef04c8c7c9457e", "occurred_at": "2026-07-11T19:45:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a4dab74dda" }, "output": { "decision": "deny", "score": 0.947, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-11, claims-fastpath v1.0 (an internally developed system) produced 'deny' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-07-11 and recorded 3 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (critical, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "critical", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.816741+00:00", "fingerprint": "4675d35ba2cc90eb8143826a" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 46.4, "reviewer_role": "adjuster", "resolved_at": "2026-07-13T18:09:53.630246+00:00" }, "ledger": { "seq": 515, "hash": "67f03f5c58104cd8e5731625d506408d04a48651ececa40c766e179d2eb314ea", "prev_hash": "7ef1295a143c89dfb6876cd27ce6f5854f9c73c9c6289abcb9ec13f8a20bf92e", "recorded_at": "2026-07-11T19:47:32+00:00" } }, { "txn_id": "BB-CLM-0442", "assessment_id": "ra_3a7d7495b53c47f0", "occurred_at": "2026-07-12T07:01:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_467f44c1db" }, "output": { "decision": "pay", "score": 0.131, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-12, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.632078+00:00", "fingerprint": "3c5fb8e30733e1f42f8166f8" }, "review": null, "ledger": { "seq": 442, "hash": "4babfe7a662c96f9c0fc06d3c56a3a3a4ffbe3e586f4157c7463ab2f5b1c47a8", "prev_hash": "f639d07b6c021a6e7847e39170fb5a0f74383b72a21e9dfb82cfa682299f36c7", "recorded_at": "2026-07-12T07:02:36+00:00" } }, { "txn_id": "BB-CLM-0524", "assessment_id": "ra_81dce762b691498e", "occurred_at": "2026-07-13T09:29:58+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_e7af7ff50c" }, "output": { "decision": "pay", "score": 0.13, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-13, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.849897+00:00", "fingerprint": "0267c8368aded8abac907f17" }, "review": null, "ledger": { "seq": 524, "hash": "eaa6a6b192f916f80570f4bd149b85a36e4a65493b576fb30d49c95dfebc8014", "prev_hash": "e65a13d4d29086887cbd0bcb344825323f360d89eb93afec4ecee98e57d373a8", "recorded_at": "2026-07-13T09:31:55+00:00" } }, { "txn_id": "BB-CLM-0501", "assessment_id": "ra_8a520f414ab94b24", "occurred_at": "2026-07-13T13:36:13+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_9158b025bb" }, "output": { "decision": "pay", "score": 0.083, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-13, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.776355+00:00", "fingerprint": "1cac64e3b063be7aa8e3933c" }, "review": null, "ledger": { "seq": 501, "hash": "1f1f390c88f22367c77eb111f6fcfb0252db94cec3590acd906cb392270b930f", "prev_hash": "c6490c95c9c5d8cafeb48389895eb8f0bae28584aad802f8ac2d9c43e9869c16", "recorded_at": "2026-07-13T13:38:10+00:00" } }, { "txn_id": "BB-CLM-0457", "assessment_id": "ra_a7ed603e94914ebf", "occurred_at": "2026-07-14T09:01:40+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_e68c05df00" }, "output": { "decision": "pay", "score": 0.038, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.664056+00:00", "fingerprint": "da995560472cce12c09140d1" }, "review": null, "ledger": { "seq": 457, "hash": "38a9bac91f134dd4e9ba841650b6766e8ab488835c102f9f47fb437a74a2a516", "prev_hash": "e53d9cf2835b1fe1792b6574e29a2706831c1510686b477b0d3ded8e8b705b2f", "recorded_at": "2026-07-14T09:04:38+00:00" } }, { "txn_id": "BB-CLM-0534", "assessment_id": "ra_52030ed1bbd74528", "occurred_at": "2026-07-14T13:04:03+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f50f7789bc" }, "output": { "decision": "deny", "score": 0.799, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-14, claims-fastpath v1.0 (an internally developed system) produced 'deny' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 2 rule(s) in force on 2026-07-14 and recorded 1 finding(s); the most serious is 'Adverse outcome from a high-risk AI system with no human review' (medium, L6).", "findings": [ { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.893871+00:00", "fingerprint": "324b5712aa77b89458a1ca7b" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 27.8, "reviewer_role": "adjuster", "resolved_at": "2026-07-15T16:53:00.417335+00:00" }, "ledger": { "seq": 534, "hash": "0381a5a4712f58b4b8f038fbafc318a3229831949966a6aafebe8c18725268e4", "prev_hash": "0fcb88c51a3193b791e3e1deaa6a10edd205673016fc005d1432c783c7e4e306", "recorded_at": "2026-07-14T13:06:41+00:00" } }, { "txn_id": "BB-CLM-0507", "assessment_id": "ra_b02db8a06fac41ef", "occurred_at": "2026-07-17T12:23:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_d65a1b63f9" }, "output": { "decision": "pay", "score": 0.198, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-17, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.793205+00:00", "fingerprint": "4eaf97d95f81523ef9355b49" }, "review": null, "ledger": { "seq": 507, "hash": "46212856b18b5eacabf6b20a6b9575890c12eb59880ddffaee04fbeb6853bb36", "prev_hash": "e30c2ec22ceb7ae8893f01f588e1750ee8878d17b2960158396a4ad235f719ca", "recorded_at": "2026-07-17T12:26:11+00:00" } }, { "txn_id": "BB-CLM-0550", "assessment_id": "ra_73be8185efd54ba7", "occurred_at": "2026-07-18T15:36:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_39e0530da4" }, "output": { "decision": "pay", "score": 0.009, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.935795+00:00", "fingerprint": "036c90058cd08a68cedef368" }, "review": null, "ledger": { "seq": 550, "hash": "f40f912923fce237eb43b0ea7d32dc522bd05c9ba84d4d6bc79c85d2461cff9c", "prev_hash": "0e9874d6fcde0636e260c2c3e2e6af2c3d3f542909c882bd33a7316e94c82db8", "recorded_at": "2026-07-18T15:39:28+00:00" } }, { "txn_id": "BB-CLM-0518", "assessment_id": "ra_37d6d352a49f4c6b", "occurred_at": "2026-07-19T02:04:24+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_04a22df1d2" }, "output": { "decision": "pay", "score": 0.186, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.829249+00:00", "fingerprint": "bd685566f30d9dd6f2b04d8b" }, "review": null, "ledger": { "seq": 518, "hash": "63f96b31b41c7b9d0f7ee99122f3a32f5a2e624f0e5a414042ad3c8585f28e8a", "prev_hash": "cb3b1e2bfaffabd72f54447297e3196f5d7ce152122bee46fe79e61eeaf1e8d6", "recorded_at": "2026-07-19T02:07:42+00:00" } }, { "txn_id": "BB-CLM-0445", "assessment_id": "ra_3b852b1069bf4cdc", "occurred_at": "2026-07-19T21:04:19+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_5dc9f58eb0" }, "output": { "decision": "pay", "score": 0.258, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.637996+00:00", "fingerprint": "4b495ce947abdd453cc8ce49" }, "review": null, "ledger": { "seq": 445, "hash": "eb54c6d14c1dca9e0cd56c21f76ddb73ff5260f6aff371bb1885527105a7156a", "prev_hash": "81823a2dfbad7c4072c03f0465bc44d34fef18f201296a7b2eb578cfbf21342d", "recorded_at": "2026-07-19T21:05:00+00:00" } }, { "txn_id": "BB-CLM-0499", "assessment_id": "ra_f9f04d8bc8ec4834", "occurred_at": "2026-07-20T21:04:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_926945382c" }, "output": { "decision": "pay", "score": 0.036, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-20, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.770185+00:00", "fingerprint": "391b4e1de826737ae1a73d00" }, "review": null, "ledger": { "seq": 499, "hash": "cc89974bc5a7f8b099b974bed1cb3171ba58883dfd50809515bc2d535baf215e", "prev_hash": "c5037cf1c9239329860b4e219cc7a4f91881905235fa667f8995f0d21f8d7627", "recorded_at": "2026-07-20T21:05:59+00:00" } }, { "txn_id": "BB-CLM-0468", "assessment_id": "ra_50a268851f3e46c2", "occurred_at": "2026-07-21T08:16:07+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f4b959d557" }, "output": { "decision": "pay", "score": 0.273, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.692604+00:00", "fingerprint": "f1125b486126898f27d9b8ad" }, "review": null, "ledger": { "seq": 468, "hash": "bc4a2cf271f4bfd6a7d131237425f1d8ea43b971074f5c3c4b167b2a8f3c1f83", "prev_hash": "893682b3af10e56499c2c8b7b92204182071151267dfa37f989ae3fb6e88d6e0", "recorded_at": "2026-07-21T08:19:44+00:00" } }, { "txn_id": "BB-CLM-0519", "assessment_id": "ra_de73d41bdf0d4979", "occurred_at": "2026-07-21T10:16:07+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_4d90b3c984" }, "output": { "decision": "deny", "score": 0.823, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, claims-fastpath v1.0 (an internally developed system) produced 'deny' in a TX auto claims transaction. Inputs included claims history, property. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.832597+00:00", "fingerprint": "60be2e1e4374ace72455dd11" }, "review": null, "ledger": { "seq": 519, "hash": "ee39c3719be86459eba94bc65aab1ddb007260e8b77fec36b0a725e2c8407965", "prev_hash": "63f96b31b41c7b9d0f7ee99122f3a32f5a2e624f0e5a414042ad3c8585f28e8a", "recorded_at": "2026-07-21T10:18:23+00:00" } }, { "txn_id": "BB-CLM-0498", "assessment_id": "ra_6fb9c33ea81740a8", "occurred_at": "2026-07-22T22:41:13+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_1f6d44aad2" }, "output": { "decision": "pay", "score": 0.17, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-07-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.768165+00:00", "fingerprint": "dc6dd79ea8b8a6bce462fb3b" }, "review": null, "ledger": { "seq": 498, "hash": "c5037cf1c9239329860b4e219cc7a4f91881905235fa667f8995f0d21f8d7627", "prev_hash": "b36b68180654b8f426a8aef07343725caa80363b0dd438d822a4292d9dfca7eb", "recorded_at": "2026-07-22T22:43:03+00:00" } }, { "txn_id": "BB-CLM-0480", "assessment_id": "ra_4239bdfabe4b48b7", "occurred_at": "2026-07-23T04:34:49+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c88c2f0888" }, "output": { "decision": "pay", "score": 0.033, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.718580+00:00", "fingerprint": "6544e71e986372f70c178c23" }, "review": null, "ledger": { "seq": 480, "hash": "527dcf2f51f9dd2151736c7ab006f18a454f1fa9b3f4be8429a9912a99c0dc56", "prev_hash": "f3d7bb66cf0662819ed6b7326675047d60bb33c1482196e5abdc9ac8cd769275", "recorded_at": "2026-07-23T04:35:41+00:00" } }, { "txn_id": "BB-CLM-0505", "assessment_id": "ra_13ab00ed942c4be0", "occurred_at": "2026-07-23T19:23:42+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_580881530b" }, "output": { "decision": "pay", "score": 0.077, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.786281+00:00", "fingerprint": "73eb779beb6c729f85488fcc" }, "review": null, "ledger": { "seq": 505, "hash": "13653e068766354c1b1bb78bed69c364472543a696cdeb59281ff3544614bafc", "prev_hash": "32811759eccc6f0943be6ce66ba517c469b48bc9a48c2e936e4597bee5f0ba6a", "recorded_at": "2026-07-23T19:25:38+00:00" } }, { "txn_id": "BB-CLM-0510", "assessment_id": "ra_5e7813365196444e", "occurred_at": "2026-07-23T19:45:59+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_6b20ef07cf" }, "output": { "decision": "pay", "score": 0.042, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.798226+00:00", "fingerprint": "3bcb2dfdf1f71461d16e91dd" }, "review": null, "ledger": { "seq": 510, "hash": "c26fcdbfc89d209c03504ab57106e6b937b8330c1bdfc9fb3c160259cb097e30", "prev_hash": "e8832cd64ce1c6aa67d9b298287e1559181fe851027d796a0d8dc014b1a0b083", "recorded_at": "2026-07-23T19:48:27+00:00" } }, { "txn_id": "BB-CLM-0479", "assessment_id": "ra_dc1dac890d294128", "occurred_at": "2026-07-25T09:45:55+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_7bb11deffe" }, "output": { "decision": "pay", "score": 0.152, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-25, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.716193+00:00", "fingerprint": "489492dd0447e529e27b18e7" }, "review": null, "ledger": { "seq": 479, "hash": "f3d7bb66cf0662819ed6b7326675047d60bb33c1482196e5abdc9ac8cd769275", "prev_hash": "88794cd0f4656361b64e1bf05aa90c49a6e38bab9e56803eba0ae74db01f85bd", "recorded_at": "2026-07-25T09:49:13+00:00" } }, { "txn_id": "BB-CLM-0452", "assessment_id": "ra_23fe8d79249d4631", "occurred_at": "2026-07-26T11:26:20+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c49fddba63" }, "output": { "decision": "pay", "score": 0.141, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-26, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.654238+00:00", "fingerprint": "2d027263a18d04d3b60b8973" }, "review": null, "ledger": { "seq": 452, "hash": "3c2995c2a4607e17e84e283126b1777ac71121e2fb2eb79ee2ed247db4f4a304", "prev_hash": "84517d9697b5806732a36b42431886420db60a40e8ae6d3d85541966d6ccbc8a", "recorded_at": "2026-07-26T11:30:12+00:00" } }, { "txn_id": "BB-CLM-0511", "assessment_id": "ra_da85628cfefa41f4", "occurred_at": "2026-07-28T17:46:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_7a37b01688" }, "output": { "decision": "pay", "score": 0.199, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-28, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.800366+00:00", "fingerprint": "d86865c6702d90cac76a880b" }, "review": null, "ledger": { "seq": 511, "hash": "febf967d2729aef3ba5cdae94dc2aeb0a71bcca2f00fc43d37ec10bc39323857", "prev_hash": "c26fcdbfc89d209c03504ab57106e6b937b8330c1bdfc9fb3c160259cb097e30", "recorded_at": "2026-07-28T17:48:54+00:00" } }, { "txn_id": "BB-CLM-0537", "assessment_id": "ra_c1bb144bd29f4f51", "occurred_at": "2026-07-29T11:16:28+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_ae440ba56b" }, "output": { "decision": "pay", "score": 0.263, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-29, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.900748+00:00", "fingerprint": "1b1bfa8bb48fd31c76e78d22" }, "review": null, "ledger": { "seq": 537, "hash": "0421ad1e6b6571f88292d4beefa05de8733e69dd7b803fe5110b33b93969205a", "prev_hash": "1063904125b6bedaebdf2a0ca279c2a7a8eca5ebd79de96fbea81f577fae7d74", "recorded_at": "2026-07-29T11:18:45+00:00" } }, { "txn_id": "BB-CLM-0471", "assessment_id": "ra_ad7f4f07734249fb", "occurred_at": "2026-07-30T17:30:34+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f1b5ded023" }, "output": { "decision": "pay", "score": 0.078, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-30, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-07-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.697492+00:00", "fingerprint": "e28039c684d0c870139a2e56" }, "review": null, "ledger": { "seq": 471, "hash": "e8ee4783f2cd73a16850d707f9a292a6fddb14b7b235e7b9922f17c3bfc53fed", "prev_hash": "b5070cfffed4829c9439b97c1557b9f5c1304827cbd3b174e0ee290355c1e544", "recorded_at": "2026-07-30T17:34:19+00:00" } }, { "txn_id": "BB-CLM-0459", "assessment_id": "ra_f797ec2ddf8b4498", "occurred_at": "2026-07-31T06:10:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_bfd02e4912" }, "output": { "decision": "pay", "score": 0.212, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.667414+00:00", "fingerprint": "29fa931a5979c17e9d55d9bc" }, "review": null, "ledger": { "seq": 459, "hash": "79cb5721e0719a6fe55573b55d80bf6cbdcab9597bb1dca96fc4e7241daca824", "prev_hash": "c5967ea4393843e0a127b5f1257fb8f7918f1bdc01d5a346ca8ed5a807e27aa6", "recorded_at": "2026-07-31T06:12:18+00:00" } }, { "txn_id": "BB-CLM-0474", "assessment_id": "ra_ffdb77416fdb449e", "occurred_at": "2026-07-31T09:34:29+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_35f2b40ad1" }, "output": { "decision": "pay", "score": 0.119, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.702950+00:00", "fingerprint": "b15037d4757ce134a3b05fe8" }, "review": null, "ledger": { "seq": 474, "hash": "e1b68730365812bcaf8276a16c6c45a3e726327df43c5fbb03fca80d253c820a", "prev_hash": "761e3b7ad6ff524a901750d452e900aa989c76dd4f7d9becca8c0bde974cf54e", "recorded_at": "2026-07-31T09:37:37+00:00" } }, { "txn_id": "BB-CLM-0541", "assessment_id": "ra_d7551e16f0d14e33", "occurred_at": "2026-08-02T00:35:37+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3a1e0c98ee" }, "output": { "decision": "pay", "score": 0.125, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-02, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.910309+00:00", "fingerprint": "76e19b1d45663b1b33ed9a07" }, "review": null, "ledger": { "seq": 541, "hash": "1e4f9daa943a2355b97464596013c31815a49d57631adc4e8bad1d55d33b1a38", "prev_hash": "0e89028b64ac04fb11d86221b10566b809ba08cb269d6d92a46f470da0fd0ce7", "recorded_at": "2026-08-02T00:36:20+00:00" } }, { "txn_id": "BB-CLM-0454", "assessment_id": "ra_3740de7e22be4cfc", "occurred_at": "2026-08-02T05:35:56+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a999e21004" }, "output": { "decision": "pay", "score": 0.195, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-02, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.659495+00:00", "fingerprint": "bdf763b080c118fbcaacac1f" }, "review": null, "ledger": { "seq": 454, "hash": "69a37bbcaf96d26d713f1c23e8a4e26b532074142655bdbb724da222f66edfdd", "prev_hash": "274a3e2c0a10bdbe4912e793f8d18c89b4750b595e5da59d8188e686aa9575ad", "recorded_at": "2026-08-02T05:38:20+00:00" } }, { "txn_id": "BB-CLM-0441", "assessment_id": "ra_b4e33220064542d5", "occurred_at": "2026-08-03T02:06:42+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_1b3dab69f6" }, "output": { "decision": "pay", "score": 0.176, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-03, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-08-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.630689+00:00", "fingerprint": "903b444d201b8779f6a251bf" }, "review": null, "ledger": { "seq": 441, "hash": "f639d07b6c021a6e7847e39170fb5a0f74383b72a21e9dfb82cfa682299f36c7", "prev_hash": "352007dcd2195f4e41be7ff432adf61a68624463c86d7adbbe21fe188a90f169", "recorded_at": "2026-08-03T02:09:05+00:00" } }, { "txn_id": "BB-CLM-0470", "assessment_id": "ra_ff40d41e6b4d43e3", "occurred_at": "2026-08-04T00:34:35+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_225c264346" }, "output": { "decision": "pay", "score": 0.261, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-04, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.696047+00:00", "fingerprint": "51de29c46c235f2501a83438" }, "review": null, "ledger": { "seq": 470, "hash": "b5070cfffed4829c9439b97c1557b9f5c1304827cbd3b174e0ee290355c1e544", "prev_hash": "e2eb52d776f6496b2ab733269273016e05cdcc01c7c8507aa3a53ca668a8b545", "recorded_at": "2026-08-04T00:35:40+00:00" } }, { "txn_id": "BB-CLM-0451", "assessment_id": "ra_7ad1d870574b416e", "occurred_at": "2026-08-06T06:11:56+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_87021b5487" }, "output": { "decision": "pay", "score": 0.061, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.652772+00:00", "fingerprint": "8ffbfacca6f7239a2211eabe" }, "review": null, "ledger": { "seq": 451, "hash": "84517d9697b5806732a36b42431886420db60a40e8ae6d3d85541966d6ccbc8a", "prev_hash": "20491702a49c13999eae15ee20dfe99d90cb1a5f24721b13eed2309183bab886", "recorded_at": "2026-08-06T06:12:45+00:00" } }, { "txn_id": "BB-CLM-0488", "assessment_id": "ra_659762b77cac4968", "occurred_at": "2026-08-06T06:35:54+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a5ce4de609" }, "output": { "decision": "pay", "score": 0.107, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.748592+00:00", "fingerprint": "d362c29e601b91fcb9bcb849" }, "review": null, "ledger": { "seq": 488, "hash": "981de397ea27a99af65a0ff7baab3b9dee7a56d3c30f5a4ec79993f1cb6dd68d", "prev_hash": "8615bc28643c4c01b4d20e5e914902b11ba9dc7005c083644a3c1cf1a518fe8e", "recorded_at": "2026-08-06T06:37:15+00:00" } }, { "txn_id": "BB-CLM-0508", "assessment_id": "ra_0695566ec7f548c3", "occurred_at": "2026-08-10T00:36:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_7386ec5bb0" }, "output": { "decision": "pay", "score": 0.08, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-10, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.795076+00:00", "fingerprint": "7efd6af932de8625d78d3912" }, "review": null, "ledger": { "seq": 508, "hash": "f548bbae22377ef61952c4e4a7c897889e72026fdeb55a3dffff61935744de58", "prev_hash": "46212856b18b5eacabf6b20a6b9575890c12eb59880ddffaee04fbeb6853bb36", "recorded_at": "2026-08-10T00:38:14+00:00" } }, { "txn_id": "BB-CLM-0483", "assessment_id": "ra_3be0d974570441c3", "occurred_at": "2026-08-10T01:47:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_27a51ef446" }, "output": { "decision": "pay", "score": 0.08, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-10, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.727975+00:00", "fingerprint": "7efd6af932de8625d78d3912" }, "review": null, "ledger": { "seq": 483, "hash": "cacef4ef633e33d37154589c6b311cfc558ce66b6a2585c87355cb03316d4221", "prev_hash": "5590ad63d2c5c5855d3ad06826919ba53fc63970c2426304d47f02baa642d1d8", "recorded_at": "2026-08-10T01:50:36+00:00" } }, { "txn_id": "BB-CLM-0504", "assessment_id": "ra_b16c1fad872748e8", "occurred_at": "2026-08-11T23:02:29+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_885a7855a2" }, "output": { "decision": "pay", "score": 0.213, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-11, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-08-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.783033+00:00", "fingerprint": "d8e4b30f09ab0f5b8112fa46" }, "review": null, "ledger": { "seq": 504, "hash": "32811759eccc6f0943be6ce66ba517c469b48bc9a48c2e936e4597bee5f0ba6a", "prev_hash": "8f3ed5649c5faa2cfcbdeb5d0b431803141883d20642be54d38ddd72261f59f8", "recorded_at": "2026-08-11T23:05:39+00:00" } }, { "txn_id": "BB-CLM-0512", "assessment_id": "ra_03b154776cb649fa", "occurred_at": "2026-08-12T06:49:37+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_4654667955" }, "output": { "decision": "pay", "score": 0.076, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-12, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.803853+00:00", "fingerprint": "901e142e5d65e2de0b51c476" }, "review": null, "ledger": { "seq": 512, "hash": "9ee56845299330b3439b19ece8bc20550daf0f8c77cb24db332c25211595998b", "prev_hash": "febf967d2729aef3ba5cdae94dc2aeb0a71bcca2f00fc43d37ec10bc39323857", "recorded_at": "2026-08-12T06:51:31+00:00" } }, { "txn_id": "BB-CLM-0506", "assessment_id": "ra_8cf128bf280a4391", "occurred_at": "2026-08-13T16:17:42+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_5821543ed1" }, "output": { "decision": "pay", "score": 0.017, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.790009+00:00", "fingerprint": "968b72f2a8102828878f6868" }, "review": null, "ledger": { "seq": 506, "hash": "e30c2ec22ceb7ae8893f01f588e1750ee8878d17b2960158396a4ad235f719ca", "prev_hash": "13653e068766354c1b1bb78bed69c364472543a696cdeb59281ff3544614bafc", "recorded_at": "2026-08-13T16:20:12+00:00" } }, { "txn_id": "BB-CLM-0489", "assessment_id": "ra_e933431dc5614b3c", "occurred_at": "2026-08-15T09:31:11+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f59e3f3a8a" }, "output": { "decision": "pay", "score": 0.026, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-15, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.750478+00:00", "fingerprint": "753a1f6aaefb57e836d578cc" }, "review": null, "ledger": { "seq": 489, "hash": "bf5b3e0ea97ffbbac8abd5fea6a62da50253de118a777f36cb144ef91c458a95", "prev_hash": "981de397ea27a99af65a0ff7baab3b9dee7a56d3c30f5a4ec79993f1cb6dd68d", "recorded_at": "2026-08-15T09:33:35+00:00" } }, { "txn_id": "BB-CLM-0532", "assessment_id": "ra_272714a1d58045a7", "occurred_at": "2026-08-18T16:54:19+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c6d3d73980" }, "output": { "decision": "pay", "score": 0.203, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.886371+00:00", "fingerprint": "a4353de3b8091431604e8dd7" }, "review": null, "ledger": { "seq": 532, "hash": "d61051d8816241137d3f279b8e6293c2e8210be36ee7e87ba9fd69fd3c72a48d", "prev_hash": "393ed46039d98f79b33a1c766d92f0f69028c73f760095522d75652fe2ad2596", "recorded_at": "2026-08-18T16:57:45+00:00" } }, { "txn_id": "BB-CLM-0449", "assessment_id": "ra_481152803d1c4acf", "occurred_at": "2026-08-18T21:36:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_74443a74c3" }, "output": { "decision": "pay", "score": 0.192, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.649292+00:00", "fingerprint": "a4353de3b8091431604e8dd7" }, "review": null, "ledger": { "seq": 449, "hash": "bfcb862b1f719f4660b4531d7a7ab6fcb6c7be2fe0efa95e6d53388e12feebce", "prev_hash": "4773ea0d41ce19f23f56d4b6e52e9049a728a0460727f5c099b957753c3927f2", "recorded_at": "2026-08-18T21:39:26+00:00" } }, { "txn_id": "BB-CLM-0496", "assessment_id": "ra_57288438c0474ade", "occurred_at": "2026-08-19T14:37:34+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_57e704c087" }, "output": { "decision": "pay", "score": 0.03, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.764758+00:00", "fingerprint": "5c819d8a579ddea395b6cac4" }, "review": null, "ledger": { "seq": 496, "hash": "89cdf625f4a8ff2decbc84d162bbfed0631665fd1b0aaa60341f6cac422ac329", "prev_hash": "053cc3ed6104a157b6b013848ba9a4d89afcd3741b1cdbacacd33f19b8a3dcb3", "recorded_at": "2026-08-19T14:39:54+00:00" } }, { "txn_id": "BB-CLM-0453", "assessment_id": "ra_0288ac6ed5d4450c", "occurred_at": "2026-08-21T01:20:26+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_18fd49a44e" }, "output": { "decision": "pay", "score": 0.095, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-21, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.657597+00:00", "fingerprint": "a9b9390bb7b7672d7fa06acf" }, "review": null, "ledger": { "seq": 453, "hash": "274a3e2c0a10bdbe4912e793f8d18c89b4750b595e5da59d8188e686aa9575ad", "prev_hash": "3c2995c2a4607e17e84e283126b1777ac71121e2fb2eb79ee2ed247db4f4a304", "recorded_at": "2026-08-21T01:22:05+00:00" } }, { "txn_id": "BB-CLM-0444", "assessment_id": "ra_c57fc923f2c04b36", "occurred_at": "2026-08-22T09:44:06+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c7166e3020" }, "output": { "decision": "pay", "score": 0.113, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.635384+00:00", "fingerprint": "df13a1c5c500ece5a6295fd2" }, "review": null, "ledger": { "seq": 444, "hash": "81823a2dfbad7c4072c03f0465bc44d34fef18f201296a7b2eb578cfbf21342d", "prev_hash": "dc1cda346d21aeda51aa9aca81619761094c535dfa811e5b951c196a9ff1fd66", "recorded_at": "2026-08-22T09:47:25+00:00" } }, { "txn_id": "BB-CLM-0535", "assessment_id": "ra_a6ed30f1ddf54384", "occurred_at": "2026-08-22T12:25:54+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_779f4efad7" }, "output": { "decision": "pay", "score": 0.055, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-08-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.895893+00:00", "fingerprint": "df86d7a7598071c0a93a43e0" }, "review": null, "ledger": { "seq": 535, "hash": "6b6f5db8292d867149f33376cd4ef4af9d3135cf994b67d8d66d1eca0df5d3c5", "prev_hash": "0381a5a4712f58b4b8f038fbafc318a3229831949966a6aafebe8c18725268e4", "recorded_at": "2026-08-22T12:28:57+00:00" } }, { "txn_id": "BB-CLM-0463", "assessment_id": "ra_0f4b2ebd7a064077", "occurred_at": "2026-08-25T07:02:42+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_04615e8612" }, "output": { "decision": "pay", "score": 0.192, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-25, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.681281+00:00", "fingerprint": "a8cbe7bc215430c832c4d5a7" }, "review": null, "ledger": { "seq": 463, "hash": "2e1022284ca0278ed8b2a09d34ab130c5feccea301a3433003527c15baa0d995", "prev_hash": "ae7d10c1cf51129f9d5fbcebf70e5ff4c89c412a9b6eb14d35531b2514807891", "recorded_at": "2026-08-25T07:06:11+00:00" } }, { "txn_id": "BB-CLM-0494", "assessment_id": "ra_5d1b5fd835754daf", "occurred_at": "2026-08-25T17:05:22+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a74f6523d4" }, "output": { "decision": "pay", "score": 0.261, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-25, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.761359+00:00", "fingerprint": "18a72edfd5153b2f9288bced" }, "review": null, "ledger": { "seq": 494, "hash": "84052fa5ef35b1d6cb43a9351e36b25fced7045549e6768187185fa2e5b3bf0a", "prev_hash": "277bf96c38dc823249a3ccdc07fb5cf49fc8de867a6ab0c5d37ac169b5148820", "recorded_at": "2026-08-25T17:08:47+00:00" } }, { "txn_id": "BB-CLM-0544", "assessment_id": "ra_2200952b089543e1", "occurred_at": "2026-08-26T05:39:39+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_29cfad3ad3" }, "output": { "decision": "pay", "score": 0.158, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-26, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.920478+00:00", "fingerprint": "65dd9a7cf723d52aa150ec56" }, "review": null, "ledger": { "seq": 544, "hash": "90c767c1820cd246fc4c7a04b4cd1e7df7806f1d604b5bffdf1b2c56c82f959b", "prev_hash": "3196120d4e8302d8cfebd958599363750573ffd6acadfeb39ce6bdc8b5bc4222", "recorded_at": "2026-08-26T05:40:49+00:00" } }, { "txn_id": "BB-CLM-0525", "assessment_id": "ra_de1f6b1fc201422f", "occurred_at": "2026-08-28T05:19:16+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_5c60ff0887" }, "output": { "decision": "pay", "score": 0.037, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-28, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.852713+00:00", "fingerprint": "1cae71fa76778f11a01e57ee" }, "review": null, "ledger": { "seq": 525, "hash": "21fad2d7f6d340201e490a09a773a5f3d9ddb8d9a5195cca778ab8c9de18e81c", "prev_hash": "eaa6a6b192f916f80570f4bd149b85a36e4a65493b576fb30d49c95dfebc8014", "recorded_at": "2026-08-28T05:21:08+00:00" } }, { "txn_id": "BB-CLM-0503", "assessment_id": "ra_1e762bc7c8eb4da1", "occurred_at": "2026-08-31T13:24:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_af0a901b22" }, "output": { "decision": "pay", "score": 0.207, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-31, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.780462+00:00", "fingerprint": "5cd02e068665085e33482279" }, "review": null, "ledger": { "seq": 503, "hash": "8f3ed5649c5faa2cfcbdeb5d0b431803141883d20642be54d38ddd72261f59f8", "prev_hash": "8fc81a3564b57467579180c064f728f7d0232455007668a70eb01e472e049483", "recorded_at": "2026-08-31T13:27:13+00:00" } }, { "txn_id": "BB-CLM-0478", "assessment_id": "ra_04d2a14542b5468c", "occurred_at": "2026-08-31T16:34:51+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c69ec203ca" }, "output": { "decision": "pay", "score": 0.131, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-31, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-08-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.714415+00:00", "fingerprint": "35683cef2e42259adac7eee8" }, "review": null, "ledger": { "seq": 478, "hash": "88794cd0f4656361b64e1bf05aa90c49a6e38bab9e56803eba0ae74db01f85bd", "prev_hash": "12f21abae400bf58a07101de2a3b335ca4364c98e8a645c5f4ade2c9cb45c015", "recorded_at": "2026-08-31T16:36:39+00:00" } }, { "txn_id": "BB-CLM-0472", "assessment_id": "ra_0c34abc504f6446d", "occurred_at": "2026-09-03T02:36:16+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_1cc0cb1777" }, "output": { "decision": "pay", "score": 0.002, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-03, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.698917+00:00", "fingerprint": "caf30b8eef45863af9288a4b" }, "review": null, "ledger": { "seq": 472, "hash": "4e6621f2b93b470f1aba1ef56add03ee268122b63e4894e95be56d2e0cd3ee3b", "prev_hash": "e8ee4783f2cd73a16850d707f9a292a6fddb14b7b235e7b9922f17c3bfc53fed", "recorded_at": "2026-09-03T02:39:12+00:00" } }, { "txn_id": "BB-CLM-0460", "assessment_id": "ra_37ea563611594053", "occurred_at": "2026-09-03T14:21:33+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_74a5b229f8" }, "output": { "decision": "partial_pay", "score": 0.582, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-09-03, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-09-03 and recorded 4 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.669300+00:00", "fingerprint": "ece378e462187ff29971423d" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 4.0, "reviewer_role": "adjuster", "resolved_at": "2026-09-03T18:25:31.352657+00:00" }, "ledger": { "seq": 460, "hash": "bee68ce7fb0d39a705708e2c227464f63c07d610dd2ed5ec722347ca19d200f9", "prev_hash": "79cb5721e0719a6fe55573b55d80bf6cbdcab9597bb1dca96fc4e7241daca824", "recorded_at": "2026-09-03T14:23:06+00:00" } }, { "txn_id": "BB-CLM-0491", "assessment_id": "ra_bb9ddfdf90724ff9", "occurred_at": "2026-09-04T20:23:03+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_992950ed4a" }, "output": { "decision": "pay", "score": 0.236, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-04, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.753847+00:00", "fingerprint": "12f38a81c3e0b0e9fd49c37e" }, "review": null, "ledger": { "seq": 491, "hash": "81817a4fe10dd8b693a12e9baa64e2242cc91c68150bd55ddff79baab849ab92", "prev_hash": "0e82d841d16d41825b7a8d019ddede300bd0c0842930fef99f396aca77c010bb", "recorded_at": "2026-09-04T20:23:57+00:00" } }, { "txn_id": "BB-CLM-0467", "assessment_id": "ra_58dee5d3b0da44b1", "occurred_at": "2026-09-05T12:40:10+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_35a8ac0f06" }, "output": { "decision": "pay", "score": 0.121, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-05, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.689617+00:00", "fingerprint": "5a81aa3db86c899b718de94f" }, "review": null, "ledger": { "seq": 467, "hash": "893682b3af10e56499c2c8b7b92204182071151267dfa37f989ae3fb6e88d6e0", "prev_hash": "a3a0a2926d3b622c8f115b87eac19b4b9c8f2bdb74b62317219c14fcaef6bf60", "recorded_at": "2026-09-05T12:42:26+00:00" } }, { "txn_id": "BB-CLM-0522", "assessment_id": "ra_bb881bc774de4908", "occurred_at": "2026-09-05T22:42:47+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_51d5b07201" }, "output": { "decision": "partial_pay", "score": 0.313, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-05, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-09-05 and recorded 3 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.843000+00:00", "fingerprint": "09fe2efbecd994c8176737dc" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 23.0, "reviewer_role": "adjuster", "resolved_at": "2026-09-06T21:48:10.803616+00:00" }, "ledger": { "seq": 522, "hash": "91dbdbc3cc0f00bdec178f24f3df0ec5b3bd2e0e81156eb7e3c458f836f395ac", "prev_hash": "1ca2864197a746a06faca1031d99fd29a3add4da944aad484e2a7b8eb749ea59", "recorded_at": "2026-09-05T22:46:30+00:00" } }, { "txn_id": "BB-CLM-0529", "assessment_id": "ra_37630b52ca8e474b", "occurred_at": "2026-09-06T09:11:12+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f7872280fe" }, "output": { "decision": "deny", "score": 0.605, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-06, claims-fastpath v1.0 (an internally developed system) produced 'deny' in a TX home claims transaction. Inputs included claims history, property. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-09-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.869063+00:00", "fingerprint": "5617425420ee095c20f40316" }, "review": null, "ledger": { "seq": 529, "hash": "445a0f7f63fabd83e85f2c69f4211ef096a5348e0c01a52b210529a1bcc2e496", "prev_hash": "e005fe9cf8e9e71fd50ce18538ae7d7a67cf9888c7a8bcd8b953e447a52e6c01", "recorded_at": "2026-09-06T09:13:00+00:00" } }, { "txn_id": "BB-CLM-0466", "assessment_id": "ra_b5cc351f48e24339", "occurred_at": "2026-09-06T19:45:41+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_74617d4b5a" }, "output": { "decision": "pay", "score": 0.111, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-06, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.686413+00:00", "fingerprint": "b7a65a46535a746d81e430be" }, "review": null, "ledger": { "seq": 466, "hash": "a3a0a2926d3b622c8f115b87eac19b4b9c8f2bdb74b62317219c14fcaef6bf60", "prev_hash": "14990ef8c1975705cad526dd5349cf1c039b476cecc18c91194fddb354d867ba", "recorded_at": "2026-09-06T19:49:03+00:00" } }, { "txn_id": "BB-CLM-0476", "assessment_id": "ra_083be0693d944b1b", "occurred_at": "2026-09-08T13:28:46+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_097c802693" }, "output": { "decision": "partial_pay", "score": 0.431, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-09-08, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-09-08 and recorded 4 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.710456+00:00", "fingerprint": "f24481bab89835b5b9c26615" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 3.7, "reviewer_role": "adjuster", "resolved_at": "2026-09-08T17:12:06.857100+00:00" }, "ledger": { "seq": 476, "hash": "bd19824e378a213793e3ca70db91091db19f037739d1e38bc719ca6dbba15e02", "prev_hash": "db4c7a8d1e4d1867687e37106c267a4e6a450e0dbff957c4a054954cec78aede", "recorded_at": "2026-09-08T13:31:17+00:00" } }, { "txn_id": "BB-CLM-0531", "assessment_id": "ra_b0819f4460e7479a", "occurred_at": "2026-09-09T03:56:06+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_fdad4cf851" }, "output": { "decision": "pay", "score": 0.065, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-09, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.882801+00:00", "fingerprint": "23224812fe54eea8b254abde" }, "review": null, "ledger": { "seq": 531, "hash": "393ed46039d98f79b33a1c766d92f0f69028c73f760095522d75652fe2ad2596", "prev_hash": "5aa5c7f6af20b0befd63f468cba92241785dcb595573625c248261ec2bc6ae57", "recorded_at": "2026-09-09T03:58:43+00:00" } }, { "txn_id": "BB-CLM-0475", "assessment_id": "ra_642f2a3086344628", "occurred_at": "2026-09-11T02:09:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a07feaeec3" }, "output": { "decision": "pay", "score": 0.034, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-11, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.706447+00:00", "fingerprint": "3daf8be8b9c44707a96990dc" }, "review": null, "ledger": { "seq": 475, "hash": "db4c7a8d1e4d1867687e37106c267a4e6a450e0dbff957c4a054954cec78aede", "prev_hash": "e1b68730365812bcaf8276a16c6c45a3e726327df43c5fbb03fca80d253c820a", "recorded_at": "2026-09-11T02:10:32+00:00" } }, { "txn_id": "BB-CLM-0443", "assessment_id": "ra_deb72d57dcc24f86", "occurred_at": "2026-09-11T06:48:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_b1c23dcd96" }, "output": { "decision": "pay", "score": 0.029, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-11, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.633795+00:00", "fingerprint": "3daf8be8b9c44707a96990dc" }, "review": null, "ledger": { "seq": 443, "hash": "dc1cda346d21aeda51aa9aca81619761094c535dfa811e5b951c196a9ff1fd66", "prev_hash": "4babfe7a662c96f9c0fc06d3c56a3a3a4ffbe3e586f4157c7463ab2f5b1c47a8", "recorded_at": "2026-09-11T06:51:45+00:00" } }, { "txn_id": "BB-CLM-0502", "assessment_id": "ra_1166b6ec3a8f4594", "occurred_at": "2026-09-11T10:19:40+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_9252b4a74e" }, "output": { "decision": "pay", "score": 0.245, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-11, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-09-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.778836+00:00", "fingerprint": "abfbeab644bf302337a40f6e" }, "review": null, "ledger": { "seq": 502, "hash": "8fc81a3564b57467579180c064f728f7d0232455007668a70eb01e472e049483", "prev_hash": "1f1f390c88f22367c77eb111f6fcfb0252db94cec3590acd906cb392270b930f", "recorded_at": "2026-09-11T10:22:37+00:00" } }, { "txn_id": "BB-CLM-0509", "assessment_id": "ra_b8ff30cc1c664f0b", "occurred_at": "2026-09-11T23:41:12+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_980a3923e9" }, "output": { "decision": "pay", "score": 0.248, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-11, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.796733+00:00", "fingerprint": "567241a19ae82bd36efd6c22" }, "review": null, "ledger": { "seq": 509, "hash": "e8832cd64ce1c6aa67d9b298287e1559181fe851027d796a0d8dc014b1a0b083", "prev_hash": "f548bbae22377ef61952c4e4a7c897889e72026fdeb55a3dffff61935744de58", "recorded_at": "2026-09-11T23:43:22+00:00" } }, { "txn_id": "BB-CLM-0526", "assessment_id": "ra_791cda461fc644c8", "occurred_at": "2026-09-13T22:54:41+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_33ada48267" }, "output": { "decision": "pay", "score": 0.255, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-13, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.859086+00:00", "fingerprint": "f8b73f7cb1d37e84ee075947" }, "review": null, "ledger": { "seq": 526, "hash": "85c12004a6969043cc074e9ca420e1f01a82791f64a6c811a0b64df1dd618dd3", "prev_hash": "21fad2d7f6d340201e490a09a773a5f3d9ddb8d9a5195cca778ab8c9de18e81c", "recorded_at": "2026-09-13T22:57:54+00:00" } }, { "txn_id": "BB-CLM-0482", "assessment_id": "ra_ca76467a78c34319", "occurred_at": "2026-09-15T03:55:36+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_ca4dbe5d3e" }, "output": { "decision": "pay", "score": 0.205, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-15, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.724113+00:00", "fingerprint": "c33e1554725a9d3dcaa8b20d" }, "review": null, "ledger": { "seq": 482, "hash": "5590ad63d2c5c5855d3ad06826919ba53fc63970c2426304d47f02baa642d1d8", "prev_hash": "bc8ed3d37d2ecc753e0f9632dbf48fb6883bdfd288edcaaeea04e3a6c9d0aaa7", "recorded_at": "2026-09-15T03:58:25+00:00" } }, { "txn_id": "BB-CLM-0520", "assessment_id": "ra_d3867f675ec740f2", "occurred_at": "2026-09-15T10:47:11+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_127bf5da36" }, "output": { "decision": "pay", "score": 0.029, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-15, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-09-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.835244+00:00", "fingerprint": "4a0b6e49904c5c3ecac9e322" }, "review": null, "ledger": { "seq": 520, "hash": "6723621619447f30b6f3eddd608833512857e2fcee1f82abe2e56af4ef115cfc", "prev_hash": "ee39c3719be86459eba94bc65aab1ddb007260e8b77fec36b0a725e2c8407965", "recorded_at": "2026-09-15T10:49:47+00:00" } }, { "txn_id": "BB-CLM-0527", "assessment_id": "ra_bfb2cc2a9dfe4a02", "occurred_at": "2026-09-18T02:05:42+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f5b1d3dbed" }, "output": { "decision": "pay", "score": 0.17, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-09-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.862174+00:00", "fingerprint": "1baf992a376c01c0e759cad3" }, "review": null, "ledger": { "seq": 527, "hash": "ec866ff09a88ee08518b696d5cc3baff6ec3e3294bf723364f34c6572ce0290d", "prev_hash": "85c12004a6969043cc074e9ca420e1f01a82791f64a6c811a0b64df1dd618dd3", "recorded_at": "2026-09-18T02:09:12+00:00" } }, { "txn_id": "BB-CLM-0486", "assessment_id": "ra_c6d519a44cca4fe1", "occurred_at": "2026-09-18T02:39:47+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c50b8b473b" }, "output": { "decision": "pay", "score": 0.041, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.741225+00:00", "fingerprint": "8165cbb28430016835a9b27d" }, "review": null, "ledger": { "seq": 486, "hash": "2cb8830921af5088b135e59af014d06edd3aef82b63c07c4f4143889e2e74cea", "prev_hash": "f1f294dc190f8b7eea86b67d483a2e22127c7ca53c41c3bb1372fddbf6d4e3e8", "recorded_at": "2026-09-18T02:42:35+00:00" } }, { "txn_id": "BB-CLM-0493", "assessment_id": "ra_8ed24270a40248b0", "occurred_at": "2026-09-21T08:21:19+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3ba43f5b09" }, "output": { "decision": "pay", "score": 0.111, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-21, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.759491+00:00", "fingerprint": "54d1806b0184579190cb9bc0" }, "review": null, "ledger": { "seq": 493, "hash": "277bf96c38dc823249a3ccdc07fb5cf49fc8de867a6ab0c5d37ac169b5148820", "prev_hash": "f212b8522fe5ee5626bebcec59f892ecf1cc5d85535756dc5f9644c33ed465af", "recorded_at": "2026-09-21T08:22:36+00:00" } }, { "txn_id": "BB-CLM-0462", "assessment_id": "ra_847cec1ca47844ce", "occurred_at": "2026-09-22T20:05:40+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_42a365ce5b" }, "output": { "decision": "pay", "score": 0.009, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-09-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.679594+00:00", "fingerprint": "80bd0e0403c8695fd363eea1" }, "review": null, "ledger": { "seq": 462, "hash": "ae7d10c1cf51129f9d5fbcebf70e5ff4c89c412a9b6eb14d35531b2514807891", "prev_hash": "cb5e1e9b0868b751d652c654fc2073cd8570a32154f28f930d83e725ff741bf4", "recorded_at": "2026-09-22T20:07:08+00:00" } }, { "txn_id": "BB-CLM-0521", "assessment_id": "ra_dd4963aab72a4df1", "occurred_at": "2026-09-24T03:21:55+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_0eb0b6a177" }, "output": { "decision": "pay", "score": 0.254, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-24, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-09-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.837817+00:00", "fingerprint": "f163fe8b1dce7dc10927f522" }, "review": null, "ledger": { "seq": 521, "hash": "1ca2864197a746a06faca1031d99fd29a3add4da944aad484e2a7b8eb749ea59", "prev_hash": "6723621619447f30b6f3eddd608833512857e2fcee1f82abe2e56af4ef115cfc", "recorded_at": "2026-09-24T03:22:43+00:00" } }, { "txn_id": "BB-CLM-0450", "assessment_id": "ra_220aa8ab08354ea6", "occurred_at": "2026-09-25T03:43:40+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_01601f9cbc" }, "output": { "decision": "pay", "score": 0.239, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-25, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.651331+00:00", "fingerprint": "fa459adf8e93af5a4649aef0" }, "review": null, "ledger": { "seq": 450, "hash": "20491702a49c13999eae15ee20dfe99d90cb1a5f24721b13eed2309183bab886", "prev_hash": "bfcb862b1f719f4660b4531d7a7ab6fcb6c7be2fe0efa95e6d53388e12feebce", "recorded_at": "2026-09-25T03:45:15+00:00" } }, { "txn_id": "BB-CLM-0540", "assessment_id": "ra_f42cde1de98c429d", "occurred_at": "2026-09-27T09:11:20+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_d2b7cc3570" }, "output": { "decision": "pay", "score": 0.161, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-27, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.908692+00:00", "fingerprint": "4c5feba9d9ebaeac9469462f" }, "review": null, "ledger": { "seq": 540, "hash": "0e89028b64ac04fb11d86221b10566b809ba08cb269d6d92a46f470da0fd0ce7", "prev_hash": "909e54648d9a95e05d722a4f27503b15776f4b9930c765addf62c6d16d044fc7", "recorded_at": "2026-09-27T09:12:46+00:00" } }, { "txn_id": "BB-CLM-0485", "assessment_id": "ra_4dc3edeb3ef34d95", "occurred_at": "2026-09-29T08:17:17+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_e08498ccf1" }, "output": { "decision": "pay", "score": 0.188, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.734882+00:00", "fingerprint": "8453d2e5bcc46bff34b95327" }, "review": null, "ledger": { "seq": 485, "hash": "f1f294dc190f8b7eea86b67d483a2e22127c7ca53c41c3bb1372fddbf6d4e3e8", "prev_hash": "916725a4a795be78b7e15b46368582a1d253978319a1457710bbfaa4c1e0b997", "recorded_at": "2026-09-29T08:19:00+00:00" } }, { "txn_id": "BB-CLM-0456", "assessment_id": "ra_4a85a003a9d74730", "occurred_at": "2026-09-29T11:05:20+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_756f133cc5" }, "output": { "decision": "pay", "score": 0.13, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.662558+00:00", "fingerprint": "1b06042d1589e8d51cf54553" }, "review": null, "ledger": { "seq": 456, "hash": "e53d9cf2835b1fe1792b6574e29a2706831c1510686b477b0d3ded8e8b705b2f", "prev_hash": "a6a437acd883c189c172bd2cfde34cfebd90e537fd47c902d0df647db4688164", "recorded_at": "2026-09-29T11:06:46+00:00" } }, { "txn_id": "BB-CLM-0448", "assessment_id": "ra_97551e5a770e4384", "occurred_at": "2026-09-30T11:58:27+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_828e74d178" }, "output": { "decision": "pay", "score": 0.051, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.647272+00:00", "fingerprint": "b92303308952d39b14105bd8" }, "review": null, "ledger": { "seq": 448, "hash": "4773ea0d41ce19f23f56d4b6e52e9049a728a0460727f5c099b957753c3927f2", "prev_hash": "9d887c94aafe5efbed5386cc38e0e178a14f51be87739cc76f5701e66acd7c17", "recorded_at": "2026-09-30T11:59:24+00:00" } }, { "txn_id": "BB-CLM-0548", "assessment_id": "ra_8eaf2a97f0294dfa", "occurred_at": "2026-09-30T21:55:46+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_167b765c14" }, "output": { "decision": "pay", "score": 0.038, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.931139+00:00", "fingerprint": "c1a51cc7dc767964180c51f3" }, "review": null, "ledger": { "seq": 548, "hash": "f2b8da26dd44faf2459cda3741d07b27d43fdb8276171e5ab3e97009d3cce29d", "prev_hash": "a80c1002b474425acb647d0e7a67fe17c87c3a242fdfd72416c0c4f73c0ce1e6", "recorded_at": "2026-09-30T21:56:30+00:00" } }, { "txn_id": "BB-CLM-0517", "assessment_id": "ra_39492dc650de49c6", "occurred_at": "2026-10-02T13:58:42+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_d51bb7c928" }, "output": { "decision": "pay", "score": 0.112, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-02, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.825857+00:00", "fingerprint": "8b56f0d02330190481933a9a" }, "review": null, "ledger": { "seq": 517, "hash": "cb3b1e2bfaffabd72f54447297e3196f5d7ce152122bee46fe79e61eeaf1e8d6", "prev_hash": "5d867ce6e4fbb7cf9eb220bfbe95691a669a3e2af9862ac1b0821f319e939ca5", "recorded_at": "2026-10-02T13:59:48+00:00" } }, { "txn_id": "BB-CLM-0484", "assessment_id": "ra_7499f6efadfe4bc1", "occurred_at": "2026-10-06T12:44:40+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_0fb12cc303" }, "output": { "decision": "pay", "score": 0.265, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-06, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.731393+00:00", "fingerprint": "24deffe8cd055b48755a6ea7" }, "review": null, "ledger": { "seq": 484, "hash": "916725a4a795be78b7e15b46368582a1d253978319a1457710bbfaa4c1e0b997", "prev_hash": "cacef4ef633e33d37154589c6b311cfc558ce66b6a2585c87355cb03316d4221", "recorded_at": "2026-10-06T12:47:55+00:00" } }, { "txn_id": "BB-CLM-0538", "assessment_id": "ra_b62efc56b3094c1c", "occurred_at": "2026-10-07T06:23:59+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_aee7001469" }, "output": { "decision": "partial_pay", "score": 0.552, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a CO home claims transaction. Inputs included claims history, property. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 2 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.903162+00:00", "fingerprint": "78795c4e99e6ef1da3aad45d" }, "review": null, "ledger": { "seq": 538, "hash": "1ea385c4f715d3a1703d0b8da029331325cf1243f0b1a44d0ab290e2a1edb8c2", "prev_hash": "0421ad1e6b6571f88292d4beefa05de8733e69dd7b803fe5110b33b93969205a", "recorded_at": "2026-10-07T06:24:53+00:00" } }, { "txn_id": "BB-CLM-0469", "assessment_id": "ra_9af2775f93ad4b9a", "occurred_at": "2026-10-07T09:24:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_1dc949a669" }, "output": { "decision": "pay", "score": 0.214, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.694059+00:00", "fingerprint": "bd4b6b95c4bb95926493b4ef" }, "review": null, "ledger": { "seq": 469, "hash": "e2eb52d776f6496b2ab733269273016e05cdcc01c7c8507aa3a53ca668a8b545", "prev_hash": "bc4a2cf271f4bfd6a7d131237425f1d8ea43b971074f5c3c4b167b2a8f3c1f83", "recorded_at": "2026-10-07T09:27:00+00:00" } }, { "txn_id": "BB-CLM-0495", "assessment_id": "ra_9c6a827409b64d7d", "occurred_at": "2026-10-07T20:06:23+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_b321b230a0" }, "output": { "decision": "pay", "score": 0.148, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.763131+00:00", "fingerprint": "4f1893317118263d79092ca2" }, "review": null, "ledger": { "seq": 495, "hash": "053cc3ed6104a157b6b013848ba9a4d89afcd3741b1cdbacacd33f19b8a3dcb3", "prev_hash": "84052fa5ef35b1d6cb43a9351e36b25fced7045549e6768187185fa2e5b3bf0a", "recorded_at": "2026-10-07T20:08:11+00:00" } }, { "txn_id": "BB-CLM-0465", "assessment_id": "ra_770b831861954513", "occurred_at": "2026-10-09T12:58:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_40c79bd6b9" }, "output": { "decision": "partial_pay", "score": 0.37, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-09, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-10-09 and recorded 4 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.684438+00:00", "fingerprint": "46e637de4600c62eadc3d465" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 7.2, "reviewer_role": "adjuster", "resolved_at": "2026-10-09T20:10:04.158254+00:00" }, "ledger": { "seq": 465, "hash": "14990ef8c1975705cad526dd5349cf1c039b476cecc18c91194fddb354d867ba", "prev_hash": "cb7b4ad1034e888e2cd3ad0d96127d289d3839ece5b9627b4ab208b837c5b0cf", "recorded_at": "2026-10-09T12:59:22+00:00" } }, { "txn_id": "BB-CLM-0539", "assessment_id": "ra_48caebd819184735", "occurred_at": "2026-10-09T19:45:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_7e67d744e7" }, "output": { "decision": "pay", "score": 0.005, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.905952+00:00", "fingerprint": "3aca9b9c52bf71e0a06f79cd" }, "review": null, "ledger": { "seq": 539, "hash": "909e54648d9a95e05d722a4f27503b15776f4b9930c765addf62c6d16d044fc7", "prev_hash": "1ea385c4f715d3a1703d0b8da029331325cf1243f0b1a44d0ab290e2a1edb8c2", "recorded_at": "2026-10-09T19:48:59+00:00" } }, { "txn_id": "BB-CLM-0516", "assessment_id": "ra_fcfe7bd9a50640f6", "occurred_at": "2026-10-11T03:52:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3f734e009f" }, "output": { "decision": "partial_pay", "score": 0.436, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-11, claims-fastpath v1.0 (an internally developed system) produced 'partial pay' in a TX auto claims transaction. Inputs included claims history, property. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-10-11 and recorded 1 finding(s); the most serious is 'Claim rejected or reduced without stated reasons' (high, L2).", "findings": [ { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.819646+00:00", "fingerprint": "9c7132a80506b99fd1c6a82c" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 12.9, "reviewer_role": "adjuster", "resolved_at": "2026-10-11T16:46:44.311710+00:00" }, "ledger": { "seq": 516, "hash": "5d867ce6e4fbb7cf9eb220bfbe95691a669a3e2af9862ac1b0821f319e939ca5", "prev_hash": "67f03f5c58104cd8e5731625d506408d04a48651ececa40c766e179d2eb314ea", "recorded_at": "2026-10-11T03:53:34+00:00" } }, { "txn_id": "BB-CLM-0542", "assessment_id": "ra_8b9bdde9bd93471a", "occurred_at": "2026-10-12T09:06:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_08822a2efc" }, "output": { "decision": "pay", "score": 0.225, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.916619+00:00", "fingerprint": "c61ff32b55a36835015aedff" }, "review": null, "ledger": { "seq": 542, "hash": "34dd9eed1d8138528f6769f5952e4d8acdd1d616db5f9aeb23cf665cf348713b", "prev_hash": "1e4f9daa943a2355b97464596013c31815a49d57631adc4e8bad1d55d33b1a38", "recorded_at": "2026-10-12T09:08:30+00:00" } }, { "txn_id": "BB-CLM-0523", "assessment_id": "ra_7f25e29673004317", "occurred_at": "2026-10-12T11:57:32+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_ca22fc5880" }, "output": { "decision": "pay", "score": 0.092, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.847178+00:00", "fingerprint": "291f5b50b21aeba643c695c8" }, "review": null, "ledger": { "seq": 523, "hash": "e65a13d4d29086887cbd0bcb344825323f360d89eb93afec4ecee98e57d373a8", "prev_hash": "91dbdbc3cc0f00bdec178f24f3df0ec5b3bd2e0e81156eb7e3c458f836f395ac", "recorded_at": "2026-10-12T12:01:08+00:00" } }, { "txn_id": "BB-CLM-0536", "assessment_id": "ra_b94a9443d7a2413d", "occurred_at": "2026-10-13T14:25:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c17a697c7b" }, "output": { "decision": "pay", "score": 0.001, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-13, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.898160+00:00", "fingerprint": "256d1e5c1da998b70f7d7887" }, "review": null, "ledger": { "seq": 536, "hash": "1063904125b6bedaebdf2a0ca279c2a7a8eca5ebd79de96fbea81f577fae7d74", "prev_hash": "6b6f5db8292d867149f33376cd4ef4af9d3135cf994b67d8d66d1eca0df5d3c5", "recorded_at": "2026-10-13T14:26:21+00:00" } }, { "txn_id": "BB-CLM-0490", "assessment_id": "ra_f5e57ed3871848ff", "occurred_at": "2026-10-14T20:27:21+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_4c3edb63d8" }, "output": { "decision": "pay", "score": 0.203, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.752073+00:00", "fingerprint": "065fab7160c4b9e69dd0b4ee" }, "review": null, "ledger": { "seq": 490, "hash": "0e82d841d16d41825b7a8d019ddede300bd0c0842930fef99f396aca77c010bb", "prev_hash": "bf5b3e0ea97ffbbac8abd5fea6a62da50253de118a777f36cb144ef91c458a95", "recorded_at": "2026-10-14T20:29:37+00:00" } }, { "txn_id": "BB-CLM-0514", "assessment_id": "ra_c29e514f987d4089", "occurred_at": "2026-10-15T02:22:24+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_d6517f215f" }, "output": { "decision": "pay", "score": 0.04, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-15, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.814517+00:00", "fingerprint": "0d9d07acfd8aafa16a0bc160" }, "review": null, "ledger": { "seq": 514, "hash": "7ef1295a143c89dfb6876cd27ce6f5854f9c73c9c6289abcb9ec13f8a20bf92e", "prev_hash": "18a1096e4556fd9f08c752c45f6cf0ebf5b75ea35b81745190699fd186b6efae", "recorded_at": "2026-10-15T02:23:41+00:00" } }, { "txn_id": "BB-CLM-0513", "assessment_id": "ra_c9d15a3559354d1d", "occurred_at": "2026-10-18T15:10:15+00:00", "state": "CO", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_8347852927" }, "output": { "decision": "pay", "score": 0.228, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-18, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 5 rule(s) in force on 2026-10-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.811146+00:00", "fingerprint": "d5dca85054638e6e4dbd3256" }, "review": null, "ledger": { "seq": 513, "hash": "18a1096e4556fd9f08c752c45f6cf0ebf5b75ea35b81745190699fd186b6efae", "prev_hash": "9ee56845299330b3439b19ece8bc20550daf0f8c77cb24db332c25211595998b", "recorded_at": "2026-10-18T15:11:36+00:00" } }, { "txn_id": "BB-CLM-0546", "assessment_id": "ra_67b36eca03c545fa", "occurred_at": "2026-10-19T07:09:20+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_e50c40f9fd" }, "output": { "decision": "pay", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-19, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-10-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.926206+00:00", "fingerprint": "9e9478c748239ba9f085aadb" }, "review": null, "ledger": { "seq": 546, "hash": "48fe3553294459d79b460fcb903fa421171b57a4e041ea51eb50070a1cdaafc5", "prev_hash": "5a7ee22fca77dc0a308bbc99f069d62ecb130c41056bb4bde412b2a4c41b4d62", "recorded_at": "2026-10-19T07:11:03+00:00" } }, { "txn_id": "BB-CLM-0464", "assessment_id": "ra_a20d729fe1ba4c33", "occurred_at": "2026-10-20T15:18:38+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_5121f011ca" }, "output": { "decision": "pay", "score": 0.149, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-20, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.682736+00:00", "fingerprint": "c319516784d3a7077c5f9fd9" }, "review": null, "ledger": { "seq": 464, "hash": "cb7b4ad1034e888e2cd3ad0d96127d289d3839ece5b9627b4ab208b837c5b0cf", "prev_hash": "2e1022284ca0278ed8b2a09d34ab130c5feccea301a3433003527c15baa0d995", "recorded_at": "2026-10-20T15:19:50+00:00" } }, { "txn_id": "BB-CLM-0477", "assessment_id": "ra_8293d8b51b3241f8", "occurred_at": "2026-10-21T18:22:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_fc1adf8df1" }, "output": { "decision": "pay", "score": 0.262, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-21, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.712660+00:00", "fingerprint": "de390234b72b59453a05385a" }, "review": null, "ledger": { "seq": 477, "hash": "12f21abae400bf58a07101de2a3b335ca4364c98e8a645c5f4ade2c9cb45c015", "prev_hash": "bd19824e378a213793e3ca70db91091db19f037739d1e38bc719ca6dbba15e02", "recorded_at": "2026-10-21T18:26:40+00:00" } }, { "txn_id": "BB-CLM-0487", "assessment_id": "ra_34ce08fc59ab41b1", "occurred_at": "2026-10-21T19:48:20+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_5dbe2f1450" }, "output": { "decision": "pay", "score": 0.139, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-21, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.745544+00:00", "fingerprint": "5b906bbb94fc9dd7b33d1254" }, "review": null, "ledger": { "seq": 487, "hash": "8615bc28643c4c01b4d20e5e914902b11ba9dc7005c083644a3c1cf1a518fe8e", "prev_hash": "2cb8830921af5088b135e59af014d06edd3aef82b63c07c4f4143889e2e74cea", "recorded_at": "2026-10-21T19:51:32+00:00" } }, { "txn_id": "BB-CLM-0455", "assessment_id": "ra_c18c307441f341e3", "occurred_at": "2026-10-22T04:46:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3ae9cd26c3" }, "output": { "decision": "pay", "score": 0.187, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.661129+00:00", "fingerprint": "c563c22b8a725db42c16e27a" }, "review": null, "ledger": { "seq": 455, "hash": "a6a437acd883c189c172bd2cfde34cfebd90e537fd47c902d0df647db4688164", "prev_hash": "69a37bbcaf96d26d713f1c23e8a4e26b532074142655bdbb724da222f66edfdd", "recorded_at": "2026-10-22T04:49:04+00:00" } }, { "txn_id": "BB-CLM-0533", "assessment_id": "ra_4b9bcdff76a14acc", "occurred_at": "2026-10-22T07:55:06+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_1cda457358" }, "output": { "decision": "pay", "score": 0.196, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.891293+00:00", "fingerprint": "c563c22b8a725db42c16e27a" }, "review": null, "ledger": { "seq": 533, "hash": "0fcb88c51a3193b791e3e1deaa6a10edd205673016fc005d1432c783c7e4e306", "prev_hash": "d61051d8816241137d3f279b8e6293c2e8210be36ee7e87ba9fd69fd3c72a48d", "recorded_at": "2026-10-22T07:59:06+00:00" } }, { "txn_id": "BB-CLM-0547", "assessment_id": "ra_0aa961e5be1046de", "occurred_at": "2026-10-22T12:09:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_a0a908a5cb" }, "output": { "decision": "pay", "score": 0.22, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.928725+00:00", "fingerprint": "c563c22b8a725db42c16e27a" }, "review": null, "ledger": { "seq": 547, "hash": "a80c1002b474425acb647d0e7a67fe17c87c3a242fdfd72416c0c4f73c0ce1e6", "prev_hash": "48fe3553294459d79b460fcb903fa421171b57a4e041ea51eb50070a1cdaafc5", "recorded_at": "2026-10-22T12:13:19+00:00" } }, { "txn_id": "BB-CLM-0497", "assessment_id": "ra_a94fb7ccea4b4392", "occurred_at": "2026-10-23T01:14:40+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_949966ca6d" }, "output": { "decision": "pay", "score": 0.027, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.766307+00:00", "fingerprint": "ea385877c9318527d5dcd225" }, "review": null, "ledger": { "seq": 497, "hash": "b36b68180654b8f426a8aef07343725caa80363b0dd438d822a4292d9dfca7eb", "prev_hash": "89cdf625f4a8ff2decbc84d162bbfed0631665fd1b0aaa60341f6cac422ac329", "recorded_at": "2026-10-23T01:16:44+00:00" } }, { "txn_id": "BB-CLM-0528", "assessment_id": "ra_6486e3c0f2a34b68", "occurred_at": "2026-10-23T07:48:38+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_fe7f1ab40f" }, "output": { "decision": "pay", "score": 0.108, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.865623+00:00", "fingerprint": "82bdc4610c3c652441fdf3d1" }, "review": null, "ledger": { "seq": 528, "hash": "e005fe9cf8e9e71fd50ce18538ae7d7a67cf9888c7a8bcd8b953e447a52e6c01", "prev_hash": "ec866ff09a88ee08518b696d5cc3baff6ec3e3294bf723364f34c6572ce0290d", "recorded_at": "2026-10-23T07:50:58+00:00" } }, { "txn_id": "BB-CLM-0473", "assessment_id": "ra_81cbf5ce13994c96", "occurred_at": "2026-10-23T13:12:39+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_3f0dee6814" }, "output": { "decision": "pay", "score": 0.189, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-23, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.700665+00:00", "fingerprint": "1289cfd78f0ea4c1d60c436f" }, "review": null, "ledger": { "seq": 473, "hash": "761e3b7ad6ff524a901750d452e900aa989c76dd4f7d9becca8c0bde974cf54e", "prev_hash": "4e6621f2b93b470f1aba1ef56add03ee268122b63e4894e95be56d2e0cd3ee3b", "recorded_at": "2026-10-23T13:15:41+00:00" } }, { "txn_id": "BB-CLM-0530", "assessment_id": "ra_bc6c91c11107441e", "occurred_at": "2026-10-25T09:42:54+00:00", "state": "TX", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_f289d039f0" }, "output": { "decision": "pay", "score": 0.102, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-25, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.877963+00:00", "fingerprint": "a19c270faa496a4194cbe1af" }, "review": null, "ledger": { "seq": 530, "hash": "5aa5c7f6af20b0befd63f468cba92241785dcb595573625c248261ec2bc6ae57", "prev_hash": "445a0f7f63fabd83e85f2c69f4211ef096a5348e0c01a52b210529a1bcc2e496", "recorded_at": "2026-10-25T09:46:03+00:00" } }, { "txn_id": "BB-CLM-0447", "assessment_id": "ra_86638bab68bc4278", "occurred_at": "2026-10-26T12:17:34+00:00", "state": "CO", "line_of_business": "home", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_20219ea7f0" }, "output": { "decision": "pay", "score": 0.283, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a CO home claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 2 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.645304+00:00", "fingerprint": "c351186aee71f0644c6fb7aa" }, "review": null, "ledger": { "seq": 447, "hash": "9d887c94aafe5efbed5386cc38e0e178a14f51be87739cc76f5701e66acd7c17", "prev_hash": "9d440619d31ff4844aad486c6deb8813caabc7a6abdc32e1e17dd467e244da3e", "recorded_at": "2026-10-26T12:18:43+00:00" } }, { "txn_id": "BB-CLM-0446", "assessment_id": "ra_6de51cd8448d4803", "occurred_at": "2026-10-27T00:30:39+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "claims-fastpath", "model_version": "1.0", "vendor": "internal", "vendor_name": null, "high_risk_class": "claims" }, "inputs": { "feature_classes": [ "claims_history", "property" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [], "consumer_token": "tok_c2541842dc" }, "output": { "decision": "pay", "score": 0.146, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-27, claims-fastpath v1.0 (an internally developed system) produced 'pay' in a TX auto claims transaction. Inputs included claims history, property. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 7 rule(s) in force on 2026-10-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.641598+00:00", "fingerprint": "d8050f044342596aeca67a44" }, "review": null, "ledger": { "seq": 446, "hash": "9d440619d31ff4844aad486c6deb8813caabc7a6abdc32e1e17dd467e244da3e", "prev_hash": "eb54c6d14c1dca9e0cd56c21f76ddb73ff5260f6aff371bb1885527105a7156a", "recorded_at": "2026-10-27T00:33:06+00:00" } } ] }, { "system_id": "fnol-fraud", "items": { "1_name_version": "Sentinel FNOL Fraud Score \u2014 versions observed: 3.1 (declared 3.1)", "2_use_cases": "Vendor model scoring first notice of loss for fraud indicators; high scores route to SIU and can hold payment. Stages: {'fraud': 118, 'claims': 2}; decisions in period: 120; adverse: 6; states: {'CO': 36, 'TX': 84}.", "3_model_type": "vendor ML ensemble", "3a_agentic": { "agentic": false, "objectives": null, "authorizations_without_human_approval": null }, "4_implementation_date": "2025-06-10", "5_development": { "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)" }, "6_purchased_modifiable": { "purchased": true, "developed_for_company": false, "modifiable": false }, "7_model_risks": { "description": "Potential adverse consumer impact evidenced by findings on the model's decisions in the period.", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "title": "Algorithmic denial, reduction or fraud hold with no human review", "severity": "critical", "layer": "L2", "count": 4 }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "severity": "high", "layer": "L4", "count": 4 }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "title": "Adverse outcome from a high-risk AI system with no human review", "severity": "medium", "layer": "L6", "count": 4 }, { "rule_id": "tx.542.rejection-without-reasons", "title": "Claim rejected or reduced without stated reasons", "severity": "high", "layer": "L2", "count": 2 } ], "adverse_decisions": 6, "dispositions": { "PASS": 116, "FLAG": 2, "HOLD": 2 } }, "8_limitations": { "carrier_documentation": "no model documentation reference recorded on the decision stream", "regbench_note": "RegBench assesses recorded facts about each decision; it does not inspect model internals or detect inputs outside the training domain." }, "9_automate_augment_support": { "declared": "automate", "observed_human_review_pct": 1.7 }, "10_testing_validation": { "testing_report_ref_on_decisions": [ "none recorded" ], "regbench_continuous_assessment": "120 of 120 decisions assessed at decision time against the rules in force on each decision date." }, "10a_proxy_evaluation": [ { "system_id": "fnol-fraud", "model_version": "3.1", "tenant_id": "bluebonnet", "state": null, "stage_mix": { "fraud": 118, "claims": 2 }, "reference_group": "A", "groups": [ { "group": "A", "decisions": 69, "adverse": 2, "adverse_rate": 0.029 }, { "group": "B", "decisions": 31, "adverse": 4, "adverse_rate": 0.129 } ], "suppressed_groups": [ "D", "C" ], "flags": [ { "group": "B", "reference": "A", "ratio": 0.897, "diff_pp": 10.0, "reasons": [ "adverse rate 10.0 points above reference" ] } ], "method": "Adverse-outcome rate by carrier-estimated demographic proxy group; reference group = largest group; adverse impact ratio = group rate / reference rate (flag below 0.80); marginal difference in percentage points (flag at 5.0 or more). Cells under 30 decisions are suppressed. Illustrative thresholds.", "computed_where": "carrier environment (proxy labels never leave it); only these aggregates are released" } ], "11_last_test_date": { "carrier": "see testing report reference", "regbench_last_assessment": "2026-10-22T18:54:58+00:00" }, "12_financial_statement_effect": "moderate", "13_legal_compliance_review": { "rules_applied": [ "doctrine.unreviewed-adverse-high-risk", "tx.542.algorithmic-denial-without-human-review", "tx.542.rejection-without-reasons", "tx.tdi.consequential-decision-without-human-review" ], "provisions_cited": [ "naic-ai-bulletin-s3-1-ais-program", "tx-ins-542-003", "tx-ins-542-055", "tx-ins-542-056", "tx-tdi-b-0003-26-governance", "tx-tdi-b-0003-26-human-review" ], "method": "Deterministic rule evaluation per decision; citations resolve against the corpus snapshot in force on the decision date or the assessment fails closed." }, "14_regulatory_actions": "None recorded in RegBench. Carrier to confirm any informal agreements, administrative complaints, consent orders or investigations." }, "case_files": [ { "txn_id": "BB-FRD-0396", "assessment_id": "ra_0c1435c0ab204249", "occurred_at": "2026-07-02T01:45:03+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_3965bcd8b2" }, "output": { "decision": "pay", "score": 0.063, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-02, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.511999+00:00", "fingerprint": "4eabf440af3e3ec5756585d7" }, "review": null, "ledger": { "seq": 396, "hash": "4c6a62202fdc21e7e64be1e30de2952426556c274cfb778256dd14a41b63174d", "prev_hash": "25292fef1df8378139704600107f01259346b3f9a52f48425251e3ecc6fa25a3", "recorded_at": "2026-07-02T01:48:30+00:00" } }, { "txn_id": "BB-FRD-0405", "assessment_id": "ra_56114dc5d2a648d8", "occurred_at": "2026-07-02T15:45:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_b84b335d8e" }, "output": { "decision": "pay", "score": 0.455, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-02, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.530888+00:00", "fingerprint": "954dad57c6076f391e30c2c3" }, "review": null, "ledger": { "seq": 405, "hash": "94396c4c8649b1452a7b281566599e1449f79463378e7869ddca446e0a67c7be", "prev_hash": "753e982d2981faab847c5a79ea83482df7640c075f6806d47e3960035160ad11", "recorded_at": "2026-07-02T15:46:51+00:00" } }, { "txn_id": "BB-FRD-0401", "assessment_id": "ra_f08b9e8faa42424d", "occurred_at": "2026-07-03T09:54:51+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_9c248907b7" }, "output": { "decision": "pay", "score": 0.104, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-03, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.523322+00:00", "fingerprint": "aec495012eac87fb3b8cdc32" }, "review": null, "ledger": { "seq": 401, "hash": "faef2f966acc312c3b12edba02e54caef1f307b1e0a5c8fae9b443cc51686ee3", "prev_hash": "7f0e994eeaf2dd4954e16d1ce169c25afbd365cca64d9073c925c5497763564d", "recorded_at": "2026-07-03T09:57:49+00:00" } }, { "txn_id": "BB-FRD-0426", "assessment_id": "ra_d61aef15a5b541b1", "occurred_at": "2026-07-06T02:04:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_598764e6f7" }, "output": { "decision": "pay", "score": 0.218, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-06, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.594427+00:00", "fingerprint": "8d29a75f5e512e2e61855de4" }, "review": null, "ledger": { "seq": 426, "hash": "14690a66f4dd5606e7cd92209fc2f11436b4515521e269b4be100335429591c2", "prev_hash": "82bacfad96c0ca75612fb772c7d6c79758b7e24989fe5dd1c3d0b664298899c3", "recorded_at": "2026-07-06T02:07:14+00:00" } }, { "txn_id": "BB-FRD-0323", "assessment_id": "ra_73201d4629de47ae", "occurred_at": "2026-07-08T19:52:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f777cd2503" }, "output": { "decision": "pay", "score": 0.358, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.332897+00:00", "fingerprint": "9db057bbdb83fcf60d8e7047" }, "review": null, "ledger": { "seq": 323, "hash": "06e1cb120247d73bc6d71f19c96b4ae998170f88420acfb17d77e948cd35b432", "prev_hash": "d862db71e02d6807eacd6f1d149accf191afd2b0ab9364e536360f48eb7bbb2a", "recorded_at": "2026-07-08T19:56:23+00:00" } }, { "txn_id": "BB-FRD-0403", "assessment_id": "ra_c0c322d0137f4230", "occurred_at": "2026-07-09T07:02:32+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_73a854da78" }, "output": { "decision": "pay", "score": 0.152, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.527557+00:00", "fingerprint": "f35ac799cf6173ec95b1220a" }, "review": null, "ledger": { "seq": 403, "hash": "2874bcd3ff65604c2e1f0995e5382e001a8a0e1bf07b3d7504d67c4299000f7d", "prev_hash": "c4a0ae3db5c4df1bcc4cb82b506ada957f63033c3e0b1bf7554769d45f3f6700", "recorded_at": "2026-07-09T07:04:25+00:00" } }, { "txn_id": "BB-FRD-0333", "assessment_id": "ra_4eb518cbb73f4037", "occurred_at": "2026-07-09T21:59:39+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_412b0ffac6" }, "output": { "decision": "pay", "score": 0.264, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.360348+00:00", "fingerprint": "62c1571f713e18561de316f0" }, "review": null, "ledger": { "seq": 333, "hash": "67f21801e10b45570b99ab61aae7c276f2fe2f881e916e94b0901f642316dd48", "prev_hash": "21aecd1322daa424ec6d8638cc207d71470a920b34964c9251c59cf0762f0792", "recorded_at": "2026-07-09T22:03:25+00:00" } }, { "txn_id": "BB-FRD-0422", "assessment_id": "ra_f5edbcf4254b4bc4", "occurred_at": "2026-07-09T22:54:38+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_2121991d4e" }, "output": { "decision": "pay", "score": 0.261, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.585263+00:00", "fingerprint": "e0edc1658d6c26203b39f473" }, "review": null, "ledger": { "seq": 422, "hash": "c61b5b19865c41f5369c8c0c129822fd5d44d7653534cb8fc8a436a0b174af09", "prev_hash": "752597e2f56c48bd51bcadf822a94bde8f475d73fb4ab873b21f18af33766a16", "recorded_at": "2026-07-09T22:55:53+00:00" } }, { "txn_id": "BB-FRD-0341", "assessment_id": "ra_db5d64f11d3a41a2", "occurred_at": "2026-07-10T09:37:58+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1e4f9c4546" }, "output": { "decision": "pay", "score": 0.338, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.385670+00:00", "fingerprint": "5bae280bcb09bf25b2a8a60f" }, "review": null, "ledger": { "seq": 341, "hash": "4b7f6553311301da25f9ac5de5c0cc34ed6ceb5ea67b55b198c11991e7ff991d", "prev_hash": "6bdc4da28dc0fc458b3f0d4a6871f3ede098ae75d059d7f41aeef7b145d22b2c", "recorded_at": "2026-07-10T09:40:30+00:00" } }, { "txn_id": "BB-FRD-0386", "assessment_id": "ra_5b516566c81041e0", "occurred_at": "2026-07-11T01:06:29+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_e3e97da2d9" }, "output": { "decision": "pay", "score": 0.13, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-11, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.489432+00:00", "fingerprint": "fad2798c5562839329e346cb" }, "review": null, "ledger": { "seq": 386, "hash": "f79ba3d8c6096f1fc9f4e015b5d0f7eef6d31328567458f66ffdd2632b840748", "prev_hash": "b159c31eba1fb665f4c6414da55ec4ec2f2120308e98bf267582e55a50f9926c", "recorded_at": "2026-07-11T01:09:31+00:00" } }, { "txn_id": "BB-FRD-0375", "assessment_id": "ra_d39f872e59c8401f", "occurred_at": "2026-07-11T02:05:48+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_475e8efe82" }, "output": { "decision": "pay", "score": 0.34, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-11, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.459946+00:00", "fingerprint": "f0d6bf41cffcc0ebbd584841" }, "review": null, "ledger": { "seq": 375, "hash": "7cebccc2072c5cf04bd2c36b859eba4f2b110ec9757e58c0f6d0da9062e4f24b", "prev_hash": "d7a469c16dde736385d245711f65250030fafa36a70d58b3575b79046f4cf9a0", "recorded_at": "2026-07-11T02:07:36+00:00" } }, { "txn_id": "BB-FRD-0402", "assessment_id": "ra_670f6010adfd4d1a", "occurred_at": "2026-07-11T13:52:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_172a958654" }, "output": { "decision": "pay", "score": 0.194, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-11, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.525623+00:00", "fingerprint": "df24fac113f7a49e213104a2" }, "review": null, "ledger": { "seq": 402, "hash": "c4a0ae3db5c4df1bcc4cb82b506ada957f63033c3e0b1bf7554769d45f3f6700", "prev_hash": "faef2f966acc312c3b12edba02e54caef1f307b1e0a5c8fae9b443cc51686ee3", "recorded_at": "2026-07-11T13:55:20+00:00" } }, { "txn_id": "BB-FRD-0324", "assessment_id": "ra_1cd57c42b6b945a2", "occurred_at": "2026-07-12T08:03:09+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_8cc1e71648" }, "output": { "decision": "pay", "score": 0.242, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-12, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.335165+00:00", "fingerprint": "c1f527d61aff37fe5c01b6b9" }, "review": null, "ledger": { "seq": 324, "hash": "3ac0db999b797e561ea716a021a8f855d08cc0736b95a4399cb74f53f2a0c389", "prev_hash": "06e1cb120247d73bc6d71f19c96b4ae998170f88420acfb17d77e948cd35b432", "recorded_at": "2026-07-12T08:05:48+00:00" } }, { "txn_id": "BB-FRD-0427", "assessment_id": "ra_1254684ab97d42ac", "occurred_at": "2026-07-13T09:15:01+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_98773fb1c6" }, "output": { "decision": "flag", "score": 0.951, "reason_codes": [ "F07" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-13, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'flag' in a TX home fraud transaction. Inputs included claims history, geography, social media. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 4 rule(s) in force on 2026-07-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.596526+00:00", "fingerprint": "bd240d074d8778c1367762bc" }, "review": null, "ledger": { "seq": 427, "hash": "880c2682029abad40f4ab861858a0503582660970733d06f7f671753cf645226", "prev_hash": "14690a66f4dd5606e7cd92209fc2f11436b4515521e269b4be100335429591c2", "recorded_at": "2026-07-13T09:15:47+00:00" } }, { "txn_id": "BB-FRD-0388", "assessment_id": "ra_f916cb2305294a92", "occurred_at": "2026-07-15T03:23:42+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_b3c8dfc4db" }, "output": { "decision": "pay", "score": 0.258, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-15, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.494451+00:00", "fingerprint": "d2519421fe4249621e903837" }, "review": null, "ledger": { "seq": 388, "hash": "601e3577323438ef2ab424cfd245501a35d0329475902bb286f8b6451614529a", "prev_hash": "fec8ba5eec8e5e456de9615d87c906ad2e039861166c3d8b013407a4552a6fd5", "recorded_at": "2026-07-15T03:27:40+00:00" } }, { "txn_id": "BB-FRD-0326", "assessment_id": "ra_6e790a57247b446f", "occurred_at": "2026-07-15T14:27:15+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_45badd340d" }, "output": { "decision": "pay", "score": 0.075, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-15, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.342301+00:00", "fingerprint": "cd440c8c87a43c6b0293851a" }, "review": null, "ledger": { "seq": 326, "hash": "ad840ae2d4e7d4824d76125ca797a014cac30ef45cdf162b473298bf859d0c65", "prev_hash": "60dc85d797fb6c6f0774de221f09d50ae2d3f3db42ff8bae21bec0ba3509ea83", "recorded_at": "2026-07-15T14:28:08+00:00" } }, { "txn_id": "BB-FRD-0382", "assessment_id": "ra_e8e89f56b5244cb2", "occurred_at": "2026-07-15T14:45:18+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_284f06dd4e" }, "output": { "decision": "pay", "score": 0.557, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-15, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.474547+00:00", "fingerprint": "298b49f1fe52398db42256b5" }, "review": null, "ledger": { "seq": 382, "hash": "cdb4d326cfc39159e9651d5bb076d002128cebf4dfc820565f7aca36f6bbb04d", "prev_hash": "8549c1830e1a009d305c719a7ed8344d879ecaf4a89cae820c6f5cf6288d514e", "recorded_at": "2026-07-15T14:49:10+00:00" } }, { "txn_id": "BB-FRD-0406", "assessment_id": "ra_5f1a6f8fae534c9d", "occurred_at": "2026-07-16T01:45:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_df3f3e08ae" }, "output": { "decision": "pay", "score": 0.264, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-16, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.532669+00:00", "fingerprint": "88c6f4cdf332eceb6c8257cc" }, "review": null, "ledger": { "seq": 406, "hash": "07740fb2083df4f9ea460427abc70bff0fadc7e75885a9e48a6089fe914894b0", "prev_hash": "94396c4c8649b1452a7b281566599e1449f79463378e7869ddca446e0a67c7be", "recorded_at": "2026-07-16T01:46:07+00:00" } }, { "txn_id": "BB-FRD-0392", "assessment_id": "ra_9187d85571574d7a", "occurred_at": "2026-07-17T21:42:23+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_22895c5120" }, "output": { "decision": "pay", "score": 0.101, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.502909+00:00", "fingerprint": "8ae26272ae3ca53ec4dec42e" }, "review": null, "ledger": { "seq": 392, "hash": "3982efb4acd84030d54977fae852c45e7f0ce75b1995edddbc1dbf8b8bac064e", "prev_hash": "f0621a6b052d2abaa6c2e4155b0c77eb2e7a47298961cb34e9c060087c52fcc8", "recorded_at": "2026-07-17T21:45:01+00:00" } }, { "txn_id": "BB-FRD-0347", "assessment_id": "ra_403b816ae18f48e8", "occurred_at": "2026-07-19T22:33:42+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_48c158511a" }, "output": { "decision": "pay", "score": 0.163, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.397724+00:00", "fingerprint": "618ab78720cc8ed84b1c10fa" }, "review": null, "ledger": { "seq": 347, "hash": "88329d71e2a18a47f3be5379492bbc94eea0b9964fcab8543cf19edf6c8fb9c2", "prev_hash": "1522a0839f7efdd207758e8179cd012fb3f16cb02f3efceb335eab5c1107b457", "recorded_at": "2026-07-19T22:37:29+00:00" } }, { "txn_id": "BB-FRD-0378", "assessment_id": "ra_18477afe82f94b14", "occurred_at": "2026-07-20T06:49:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_4a6f50c540" }, "output": { "decision": "pay", "score": 0.342, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.464982+00:00", "fingerprint": "57f608e9d9e432f7779ee8f7" }, "review": null, "ledger": { "seq": 378, "hash": "ef4cf3aac6fa1f3eef6486aba504602fcf143e1055fd7ef9090166fc2a649145", "prev_hash": "a80f4d005d4bb7a219153a7938b7d3b183f5ebbad6ac170f6e520cecf618fa86", "recorded_at": "2026-07-20T06:50:43+00:00" } }, { "txn_id": "BB-FRD-0365", "assessment_id": "ra_fbb573be214a4008", "occurred_at": "2026-07-20T07:17:18+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f7f4f87c2b" }, "output": { "decision": "pay", "score": 0.508, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.439138+00:00", "fingerprint": "d7ad41fe87089fba2870a9d4" }, "review": null, "ledger": { "seq": 365, "hash": "0bff4c578d36d68826428e5dfc90691f077b34ed2e376703cee1c1f954d52a80", "prev_hash": "1e2b17db41528b0c93da22446e3989bad1d982fee9ed17bcb72c4d597b381edb", "recorded_at": "2026-07-20T07:19:37+00:00" } }, { "txn_id": "BB-FRD-0413", "assessment_id": "ra_f9dde2502cbb492d", "occurred_at": "2026-07-20T14:33:50+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_5568957a78" }, "output": { "decision": "pay", "score": 0.407, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.564893+00:00", "fingerprint": "714430fd2cdc61b4731df053" }, "review": null, "ledger": { "seq": 413, "hash": "40cb1bf69fc9f2ac0d89814616ab6a1dfcfe38ca628cc3d4a8f192f1a945eae8", "prev_hash": "e9143fd6976a036c090b371c67a7724ba55cfbd7687855149a40b3f93b2efe4f", "recorded_at": "2026-07-20T14:36:04+00:00" } }, { "txn_id": "BB-FRD-0335", "assessment_id": "ra_71073f6567c54159", "occurred_at": "2026-07-21T02:49:22+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1957306892" }, "output": { "decision": "pay", "score": 0.229, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.369043+00:00", "fingerprint": "4a45dcef91989cb53dc38f74" }, "review": null, "ledger": { "seq": 335, "hash": "41776227e405b64d9df6ddfb5d271582169e13b8f69a9c7f406ac2f2e8dbd483", "prev_hash": "888482c0dbb5d84b721bad414b005d2dda75b91c45f607ffe94098e5d8b2a5e8", "recorded_at": "2026-07-21T02:51:53+00:00" } }, { "txn_id": "BB-FRD-0343", "assessment_id": "ra_dda20bbf39264f01", "occurred_at": "2026-07-21T07:30:39+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_b153f502a3" }, "output": { "decision": "pay", "score": 0.666, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.390732+00:00", "fingerprint": "aca29f9c8c7de565ed55cd83" }, "review": null, "ledger": { "seq": 343, "hash": "2b6da68f27319a26f6fd2c90bb9f34cdf6c35841d8067177e0a5b659f06ccf94", "prev_hash": "89667068b1213bc7a60d4f97264050f2a40087e0f754aa392ce5c5373a27b5df", "recorded_at": "2026-07-21T07:31:20+00:00" } }, { "txn_id": "BB-FRD-0390", "assessment_id": "ra_367a6a8f04f64004", "occurred_at": "2026-07-21T16:03:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_cb504f1c8b" }, "output": { "decision": "pay", "score": 0.273, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.498423+00:00", "fingerprint": "6c06e309637ffd183a920971" }, "review": null, "ledger": { "seq": 390, "hash": "ac0a9ed479bb4cece5bbfdd5f7dd17f7ce6238f10200393364556d21d980beb9", "prev_hash": "9bf89fe4d8d5278874b90fabd34c064f56f972860f6c8ae643ef538a8511b8ab", "recorded_at": "2026-07-21T16:06:33+00:00" } }, { "txn_id": "BB-FRD-0329", "assessment_id": "ra_e40db48e08c64eef", "occurred_at": "2026-07-22T23:15:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_d160dca1e1" }, "output": { "decision": "pay", "score": 0.219, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-22, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.349204+00:00", "fingerprint": "def212028e863d37d0eb14ab" }, "review": null, "ledger": { "seq": 329, "hash": "516a1a4b876a558dacfd6359089e62a8b028b99c816024c20ece6b0e69bdbafe", "prev_hash": "2ad7b6b38ad5eb56bd9970f4d6d74a9dc10c9970652949b6975d86c601d5a272", "recorded_at": "2026-07-22T23:18:27+00:00" } }, { "txn_id": "BB-FRD-0399", "assessment_id": "ra_7880366e727a46eb", "occurred_at": "2026-07-23T10:54:05+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_37523b4e27" }, "output": { "decision": "pay", "score": 0.283, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.517899+00:00", "fingerprint": "be54a09fab400c6a3852d9f0" }, "review": null, "ledger": { "seq": 399, "hash": "6ca2e53d1e5cbd34f5d4d3c2fc78199746c37a90896f94bf9430dae8caa847b8", "prev_hash": "480c552153524ee68ff6da941072daead12c1a673ac0cc6d268ade19db0d0293", "recorded_at": "2026-07-23T10:55:20+00:00" } }, { "txn_id": "BB-FRD-0420", "assessment_id": "ra_4dcdbe64a9cb49af", "occurred_at": "2026-07-23T17:41:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_c48a42d18f" }, "output": { "decision": "pay", "score": 0.027, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.582042+00:00", "fingerprint": "4919dd930b31478225c6fdf1" }, "review": null, "ledger": { "seq": 420, "hash": "4de2bee307ca34a71280825ef9d54a319379774bee5c10a6776e18ac5ed6d1d8", "prev_hash": "99742716f189d1ad79905283e1b40cf994cc24eefcedc94acac1e0495202765e", "recorded_at": "2026-07-23T17:43:19+00:00" } }, { "txn_id": "BB-FRD-0438", "assessment_id": "ra_d95dc62ceab94fc0", "occurred_at": "2026-07-27T11:17:35+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_45bb6bc170" }, "output": { "decision": "pay", "score": 0.042, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-27, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.624093+00:00", "fingerprint": "be117ca7332a22da81e58c36" }, "review": null, "ledger": { "seq": 438, "hash": "93c24ff920cc36260841fb70114fedf956f3a5a73187f12a00923abb236a0453", "prev_hash": "2cca4f2db462640888a9691be667d4a72e98c19c53634ae7070f0f903692da84", "recorded_at": "2026-07-27T11:18:19+00:00" } }, { "txn_id": "BB-FRD-0334", "assessment_id": "ra_bb0187e44db04815", "occurred_at": "2026-07-27T14:14:20+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_ec552cc227" }, "output": { "decision": "pay", "score": 0.144, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-27, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.363812+00:00", "fingerprint": "40523df1ecd30f61ba194649" }, "review": null, "ledger": { "seq": 334, "hash": "888482c0dbb5d84b721bad414b005d2dda75b91c45f607ffe94098e5d8b2a5e8", "prev_hash": "67f21801e10b45570b99ab61aae7c276f2fe2f881e916e94b0901f642316dd48", "recorded_at": "2026-07-27T14:17:03+00:00" } }, { "txn_id": "BB-FRD-0370", "assessment_id": "ra_e8ccd81d8b774157", "occurred_at": "2026-07-28T04:08:36+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_4981636e39" }, "output": { "decision": "pay", "score": 0.086, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-28, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.449236+00:00", "fingerprint": "7e27782407f486e7f7ec15ad" }, "review": null, "ledger": { "seq": 370, "hash": "f7d812923644bd8516586af63d481f977fe343a2fccb91c18393bdf8ff809f66", "prev_hash": "867c469328cf326604c4cb7b9b2684660dc052a409ff1101db260c301b483ae7", "recorded_at": "2026-07-28T04:11:39+00:00" } }, { "txn_id": "BB-FRD-0366", "assessment_id": "ra_f0fa17adb5a24033", "occurred_at": "2026-07-30T01:18:59+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f5c87328d8" }, "output": { "decision": "pay", "score": 0.257, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-30, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-07-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.442478+00:00", "fingerprint": "90bcf8a53e6327ad4eb8733e" }, "review": null, "ledger": { "seq": 366, "hash": "a39f0f12e51d030f0ba072395ae46b931259cf44101e51e291cdae6fbd4b0149", "prev_hash": "0bff4c578d36d68826428e5dfc90691f077b34ed2e376703cee1c1f954d52a80", "recorded_at": "2026-07-30T01:19:39+00:00" } }, { "txn_id": "BB-FRD-0344", "assessment_id": "ra_4e0c8082531943ad", "occurred_at": "2026-07-31T03:51:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_a707481377" }, "output": { "decision": "pay", "score": 0.324, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.392527+00:00", "fingerprint": "f67d9fc88b20b8123baef0b6" }, "review": null, "ledger": { "seq": 344, "hash": "d7c0e3f4abbcc3a42f1f9fa18ed4cae6f252f65005b95ae6f7cadac06924474b", "prev_hash": "2b6da68f27319a26f6fd2c90bb9f34cdf6c35841d8067177e0a5b659f06ccf94", "recorded_at": "2026-07-31T03:52:11+00:00" } }, { "txn_id": "BB-FRD-0391", "assessment_id": "ra_4c9f96f8951e45b0", "occurred_at": "2026-07-31T22:01:41+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_9bc4ee3d0f" }, "output": { "decision": "pay", "score": 0.386, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.501076+00:00", "fingerprint": "86d105d5bc3d418b300c7047" }, "review": null, "ledger": { "seq": 391, "hash": "f0621a6b052d2abaa6c2e4155b0c77eb2e7a47298961cb34e9c060087c52fcc8", "prev_hash": "ac0a9ed479bb4cece5bbfdd5f7dd17f7ce6238f10200393364556d21d980beb9", "recorded_at": "2026-07-31T22:02:30+00:00" } }, { "txn_id": "BB-FRD-0387", "assessment_id": "ra_5b7411ee704247da", "occurred_at": "2026-08-01T02:05:58+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1f78e88cf3" }, "output": { "decision": "pay", "score": 0.559, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-01, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-01 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.492892+00:00", "fingerprint": "fb44c1abe158521bdf09c3c0" }, "review": null, "ledger": { "seq": 387, "hash": "fec8ba5eec8e5e456de9615d87c906ad2e039861166c3d8b013407a4552a6fd5", "prev_hash": "f79ba3d8c6096f1fc9f4e015b5d0f7eef6d31328567458f66ffdd2632b840748", "recorded_at": "2026-08-01T02:08:06+00:00" } }, { "txn_id": "BB-FRD-0371", "assessment_id": "ra_497019e40a5f4a95", "occurred_at": "2026-08-01T05:24:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_c10a1ce866" }, "output": { "decision": "pay", "score": 0.325, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-01, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-01 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.450872+00:00", "fingerprint": "2627cf3d367d6dc2bc4cf8e5" }, "review": null, "ledger": { "seq": 371, "hash": "519b6789c4a9e688582e7d32ffcab3daadff1098ac1e5ffeba2137d0f01b58a7", "prev_hash": "f7d812923644bd8516586af63d481f977fe343a2fccb91c18393bdf8ff809f66", "recorded_at": "2026-08-01T05:28:45+00:00" } }, { "txn_id": "BB-FRD-0367", "assessment_id": "ra_13081f3e4b5645b7", "occurred_at": "2026-08-03T10:34:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_fef8d0f97e" }, "output": { "decision": "pay", "score": 0.298, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-03, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.444137+00:00", "fingerprint": "d470fbd0881a733bf1a0c08b" }, "review": null, "ledger": { "seq": 367, "hash": "9963bc2fb0b6194da098f87080845d7422042e2e758df046508f2b9d9c3d8bae", "prev_hash": "a39f0f12e51d030f0ba072395ae46b931259cf44101e51e291cdae6fbd4b0149", "recorded_at": "2026-08-03T10:35:34+00:00" } }, { "txn_id": "BB-FRD-0394", "assessment_id": "ra_f62d8ae56e0e4c01", "occurred_at": "2026-08-03T21:44:14+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_d33c019b9a" }, "output": { "decision": "pay", "score": 0.088, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-03, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.508402+00:00", "fingerprint": "6a1897d209fd87a4b43c404c" }, "review": null, "ledger": { "seq": 394, "hash": "3a80c1fc7cfee26dca115b62c082db1d61ac178010afb04f52f80963396e5682", "prev_hash": "c02670c102b53fa9c7db9d03beaede6acea3a304b26acec20b019a62ace58797", "recorded_at": "2026-08-03T21:45:05+00:00" } }, { "txn_id": "BB-FRD-0339", "assessment_id": "ra_31e8c870c38b4d91", "occurred_at": "2026-08-05T14:33:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_61928de894" }, "output": { "decision": "pay", "score": 0.208, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.381059+00:00", "fingerprint": "7bdc3eb615d9324c30013583" }, "review": null, "ledger": { "seq": 339, "hash": "04158e61d7d961c129196e2cbe73720033a73c220bef00b2dfd27ebde84472a0", "prev_hash": "e7676d1afa49a117d740ffcee9e3fe92cc036de71378f631023e91bebc8b1dd8", "recorded_at": "2026-08-05T14:34:18+00:00" } }, { "txn_id": "BB-FRD-0421", "assessment_id": "ra_ed959c4688fd4496", "occurred_at": "2026-08-06T18:41:28+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_5edadc492d" }, "output": { "decision": "pay", "score": 0.227, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.583680+00:00", "fingerprint": "acb457930ff987db0ccc8fc6" }, "review": null, "ledger": { "seq": 421, "hash": "752597e2f56c48bd51bcadf822a94bde8f475d73fb4ab873b21f18af33766a16", "prev_hash": "4de2bee307ca34a71280825ef9d54a319379774bee5c10a6776e18ac5ed6d1d8", "recorded_at": "2026-08-06T18:44:08+00:00" } }, { "txn_id": "BB-FRD-0398", "assessment_id": "ra_13a00b7a2445404f", "occurred_at": "2026-08-07T22:57:55+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_d051ecd206" }, "output": { "decision": "pay", "score": 0.419, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-07, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.515779+00:00", "fingerprint": "bf734c45a44274d272f722fb" }, "review": null, "ledger": { "seq": 398, "hash": "480c552153524ee68ff6da941072daead12c1a673ac0cc6d268ade19db0d0293", "prev_hash": "e3d4e4e746334022ab5b3360a86e661b9bafbdbf0a5d75828acd870e291e0ce5", "recorded_at": "2026-08-07T23:01:20+00:00" } }, { "txn_id": "BB-FRD-0353", "assessment_id": "ra_b2addcdf736945b5", "occurred_at": "2026-08-08T09:34:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_fe85ae0309" }, "output": { "decision": "pay", "score": 0.246, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.412075+00:00", "fingerprint": "22ff433ba12e207e0a45f77f" }, "review": null, "ledger": { "seq": 353, "hash": "c7184c789a780b4685e12020784f54ebc4bef9367ecbe951750095a79164556b", "prev_hash": "55af45b3a934c5ea6ddab360371ca4ab8d0e03e37af525a0c2e9d085c0eb6aae", "recorded_at": "2026-08-08T09:35:47+00:00" } }, { "txn_id": "BB-FRD-0433", "assessment_id": "ra_2d7a4bdf316a4953", "occurred_at": "2026-08-08T10:15:12+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_92d0c2acb2" }, "output": { "decision": "pay", "score": 0.175, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.610320+00:00", "fingerprint": "13715e10d40ac96acf9c7cc7" }, "review": null, "ledger": { "seq": 433, "hash": "eb44426761585b70e770ad530f9e6929a73fe0a8b48e0de02bb4f84f9dcacd14", "prev_hash": "7179f61f54dcccdc968958b353db6080c0af7f7843453e40425d57d01b93e2cb", "recorded_at": "2026-08-08T10:16:14+00:00" } }, { "txn_id": "BB-FRD-0425", "assessment_id": "ra_ca418ecfd8b249ea", "occurred_at": "2026-08-08T15:53:43+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1a51deb076" }, "output": { "decision": "pay", "score": 0.229, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.592809+00:00", "fingerprint": "398fcdf35884076b7050b12c" }, "review": null, "ledger": { "seq": 425, "hash": "82bacfad96c0ca75612fb772c7d6c79758b7e24989fe5dd1c3d0b664298899c3", "prev_hash": "4f1c5a3de37738a086a59b2bf7af80686b63d71f3dec97a1c344d399b5b435ec", "recorded_at": "2026-08-08T15:57:40+00:00" } }, { "txn_id": "BB-FRD-0377", "assessment_id": "ra_cb497de0c0f4424d", "occurred_at": "2026-08-11T01:45:35+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f61ba52f2a" }, "output": { "decision": "pay", "score": 0.115, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-11, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.463444+00:00", "fingerprint": "45ecd46165a2250015b80a49" }, "review": null, "ledger": { "seq": 377, "hash": "a80f4d005d4bb7a219153a7938b7d3b183f5ebbad6ac170f6e520cecf618fa86", "prev_hash": "3627327df442825095448eb94cc93b6c2887416165b2b8483df3257051c00f8e", "recorded_at": "2026-08-11T01:46:30+00:00" } }, { "txn_id": "BB-FRD-0362", "assessment_id": "ra_a15a68e0d89c4bf9", "occurred_at": "2026-08-13T23:00:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_892b3afd71" }, "output": { "decision": "pay", "score": 0.245, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.432459+00:00", "fingerprint": "c28a86f75d879c7decfe4f5d" }, "review": null, "ledger": { "seq": 362, "hash": "9479c6d8bfdc12f12d30ff79f38ae5bb5e827ae6ea3fd1c2f3f486344903e32d", "prev_hash": "6c531e5a70cef257dcf810d888d71115f397b76a717f70e3eb950aa4f60ef8b9", "recorded_at": "2026-08-13T23:01:58+00:00" } }, { "txn_id": "BB-FRD-0351", "assessment_id": "ra_0e1bb59194a44395", "occurred_at": "2026-08-14T05:21:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_477757496d" }, "output": { "decision": "flag", "score": 0.864, "reason_codes": [ "F07" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'flag' in a TX auto fraud transaction. Inputs included claims history, geography, social media. A human reviewed the output by a adjuster. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 4 rule(s) in force on 2026-08-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.407352+00:00", "fingerprint": "9f91d1a22755f42a443f3aa3" }, "review": null, "ledger": { "seq": 351, "hash": "791c3272fbd8e9177c66cf71a90dc6629870709673ae0ee2287c2615707d116b", "prev_hash": "add1a606c6f45f64d2be7411806e8cb2aa67164173bef2b0dae9720536c17953", "recorded_at": "2026-08-14T05:23:05+00:00" } }, { "txn_id": "BB-FRD-0358", "assessment_id": "ra_dcd6191242e04c62", "occurred_at": "2026-08-15T16:28:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_4363305db7" }, "output": { "decision": "flag", "score": 0.816, "reason_codes": [ "F07" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-15, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'flag' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 4 rule(s) in force on 2026-08-15 and recorded 3 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.424676+00:00", "fingerprint": "0ff41b63bbdfedcda463d0ae" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 358, "hash": "3562279c9896ff8d1d678897a7bf593b6d86ac263ec6aec72875b4f20dfdf63c", "prev_hash": "c383f83a24480a30ba0140e53481c7cd07f23eb0fa6fb16fe226cab488f41067", "recorded_at": "2026-08-15T16:30:35+00:00" } }, { "txn_id": "BB-FRD-0393", "assessment_id": "ra_e7586fa0b48d49cc", "occurred_at": "2026-08-16T12:38:33+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_d2d47681b4" }, "output": { "decision": "pay", "score": 0.183, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-16, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.506143+00:00", "fingerprint": "2c6344bb50754522c4a2a1f6" }, "review": null, "ledger": { "seq": 393, "hash": "c02670c102b53fa9c7db9d03beaede6acea3a304b26acec20b019a62ace58797", "prev_hash": "3982efb4acd84030d54977fae852c45e7f0ce75b1995edddbc1dbf8b8bac064e", "recorded_at": "2026-08-16T12:39:46+00:00" } }, { "txn_id": "BB-FRD-0340", "assessment_id": "ra_6c3f940baf414fb4", "occurred_at": "2026-08-16T18:31:54+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_904f07f59b" }, "output": { "decision": "pay", "score": 0.268, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-16, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.382984+00:00", "fingerprint": "03a25dee1d771fdd10e9ae04" }, "review": null, "ledger": { "seq": 340, "hash": "6bdc4da28dc0fc458b3f0d4a6871f3ede098ae75d059d7f41aeef7b145d22b2c", "prev_hash": "04158e61d7d961c129196e2cbe73720033a73c220bef00b2dfd27ebde84472a0", "recorded_at": "2026-08-16T18:34:42+00:00" } }, { "txn_id": "BB-FRD-0346", "assessment_id": "ra_77e8d5525a9f4641", "occurred_at": "2026-08-17T23:55:44+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_b1b93d6639" }, "output": { "decision": "pay", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.395600+00:00", "fingerprint": "df0229931f53ccaf4abcecaf" }, "review": null, "ledger": { "seq": 346, "hash": "1522a0839f7efdd207758e8179cd012fb3f16cb02f3efceb335eab5c1107b457", "prev_hash": "0510035612beed985551e6b8fd7dc44177dfd384fb47a0cdf8149b366dfb7b7e", "recorded_at": "2026-08-17T23:57:30+00:00" } }, { "txn_id": "BB-FRD-0430", "assessment_id": "ra_f4cb70d7137d40bc", "occurred_at": "2026-08-18T17:32:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_b68d58b8c0" }, "output": { "decision": "pay", "score": 0.115, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-18, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.602044+00:00", "fingerprint": "18655bcb1c1143625ecbb6e3" }, "review": null, "ledger": { "seq": 430, "hash": "26f5c7353fa09839ff7a872b0e13e82b963765b4935a6e3337fb367ded6335e8", "prev_hash": "2b537b61d5b726629f9588098661e15b66d2080edbc77a17383880a40f552377", "recorded_at": "2026-08-18T17:35:01+00:00" } }, { "txn_id": "BB-FRD-0397", "assessment_id": "ra_7945742011be4274", "occurred_at": "2026-08-19T04:08:05+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_303997a2e3" }, "output": { "decision": "pay", "score": 0.113, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.514127+00:00", "fingerprint": "792c8d8f3fe2cc7491bd6cb6" }, "review": null, "ledger": { "seq": 397, "hash": "e3d4e4e746334022ab5b3360a86e661b9bafbdbf0a5d75828acd870e291e0ce5", "prev_hash": "4c6a62202fdc21e7e64be1e30de2952426556c274cfb778256dd14a41b63174d", "recorded_at": "2026-08-19T04:09:12+00:00" } }, { "txn_id": "BB-FRD-0379", "assessment_id": "ra_e67f05228d094f79", "occurred_at": "2026-08-19T07:07:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_ffa181ea9c" }, "output": { "decision": "pay", "score": 0.053, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.466552+00:00", "fingerprint": "8b11d970b8329845209e538c" }, "review": null, "ledger": { "seq": 379, "hash": "8dc3e087bdfdfb869f3f9e007ddd85778bd39bed888431c687237ef99f154697", "prev_hash": "ef4cf3aac6fa1f3eef6486aba504602fcf143e1055fd7ef9090166fc2a649145", "recorded_at": "2026-08-19T07:09:13+00:00" } }, { "txn_id": "BB-FRD-0357", "assessment_id": "ra_80055aeedad040b6", "occurred_at": "2026-08-20T16:54:46+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_a8d90dcdf0" }, "output": { "decision": "pay", "score": 0.181, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.420311+00:00", "fingerprint": "ac39e808c9b19abda516476c" }, "review": null, "ledger": { "seq": 357, "hash": "c383f83a24480a30ba0140e53481c7cd07f23eb0fa6fb16fe226cab488f41067", "prev_hash": "1d2e06cfb11af029f91a3c2b769f935e56320a9243317b0caf29837ebcda4b83", "recorded_at": "2026-08-20T16:56:24+00:00" } }, { "txn_id": "BB-FRD-0411", "assessment_id": "ra_40f16f042b8948f2", "occurred_at": "2026-08-20T23:14:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_2300dc6936" }, "output": { "decision": "pay", "score": 0.381, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.558584+00:00", "fingerprint": "a03ef7dc8d2a1eb2f99c95c3" }, "review": null, "ledger": { "seq": 411, "hash": "f747d9c78b7bba09560edb7d605ad78e6dfb8fb8887720f3762f0cb2d9b73cc2", "prev_hash": "8b5a7e6d69a01e00dfb3b2538ee8f48914104150926b2e88c1c2598e3666b745", "recorded_at": "2026-08-20T23:18:35+00:00" } }, { "txn_id": "BB-FRD-0360", "assessment_id": "ra_3dcb471187954449", "occurred_at": "2026-08-21T03:23:17+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_dc7c2c3d9a" }, "output": { "decision": "pay", "score": 0.297, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.429258+00:00", "fingerprint": "f8776d9c8c7e7bae461fe910" }, "review": null, "ledger": { "seq": 360, "hash": "1cd3ae48fa25be5da2e7c2cbe7a30a3ccc549a85b99555a60f2e97b2b2bc2de2", "prev_hash": "7bb2dc3dbb5db473bda98dcc67b79c8a6677a0b673fb79d2648f5fb61d4433a2", "recorded_at": "2026-08-21T03:24:30+00:00" } }, { "txn_id": "BB-FRD-0409", "assessment_id": "ra_53ac84b8c94b4ec1", "occurred_at": "2026-08-22T12:04:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_fbd26d945c" }, "output": { "decision": "pay", "score": 0.176, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-22, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.546916+00:00", "fingerprint": "4e708fc9d6129823738fc86d" }, "review": null, "ledger": { "seq": 409, "hash": "cbcdda97f09c6c76ebf2a64034c9e205b1265be7cf265dbae09ec6cd535d14a8", "prev_hash": "770bc6f4deb31fe6d4580cc4cbae96187016221ef6ee17425044fd68f12ad770", "recorded_at": "2026-08-22T12:06:18+00:00" } }, { "txn_id": "BB-FRD-0348", "assessment_id": "ra_8d7ea8c2bfcd4010", "occurred_at": "2026-08-22T13:16:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f67cdf9d11" }, "output": { "decision": "pay", "score": 0.049, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-22, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-08-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.399687+00:00", "fingerprint": "a16671e43a843a42168c15e9" }, "review": null, "ledger": { "seq": 348, "hash": "9dc846aceee2458f567486d67eeadd813f8335416eae800f2ce50d73bb3d5ca9", "prev_hash": "88329d71e2a18a47f3be5379492bbc94eea0b9964fcab8543cf19edf6c8fb9c2", "recorded_at": "2026-08-22T13:19:25+00:00" } }, { "txn_id": "BB-FRD-0368", "assessment_id": "ra_2cdf59d730bf4280", "occurred_at": "2026-08-24T02:29:18+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_5765598ec9" }, "output": { "decision": "pay", "score": 0.316, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-24, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.445689+00:00", "fingerprint": "a8887aafcc4c7d35a0be0823" }, "review": null, "ledger": { "seq": 368, "hash": "29391975e8d05533a0360aeb54159358bca8bdd39f4a517912c5a00dcb66eaad", "prev_hash": "9963bc2fb0b6194da098f87080845d7422042e2e758df046508f2b9d9c3d8bae", "recorded_at": "2026-08-24T02:30:56+00:00" } }, { "txn_id": "BB-FRD-0410", "assessment_id": "ra_ad1f062f57b24738", "occurred_at": "2026-08-29T19:26:05+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_45c63482c2" }, "output": { "decision": "pay", "score": 0.407, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-29, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.551316+00:00", "fingerprint": "045330ad02b6404e1536eb01" }, "review": null, "ledger": { "seq": 410, "hash": "8b5a7e6d69a01e00dfb3b2538ee8f48914104150926b2e88c1c2598e3666b745", "prev_hash": "cbcdda97f09c6c76ebf2a64034c9e205b1265be7cf265dbae09ec6cd535d14a8", "recorded_at": "2026-08-29T19:28:35+00:00" } }, { "txn_id": "BB-FRD-0373", "assessment_id": "ra_58185b3d52244c11", "occurred_at": "2026-08-30T16:19:25+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_e5f876b3e9" }, "output": { "decision": "pay", "score": 0.678, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-30, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-08-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.454439+00:00", "fingerprint": "7cab1b4677ea20136217779f" }, "review": null, "ledger": { "seq": 373, "hash": "98afa7f0489ca3870049b7d8f63237fb128150a1ea72fffc8d8197b235dea8b6", "prev_hash": "94328e8be38b94a1a0407274a95d233f2c83fdaf64cf329b96ce017266fa608b", "recorded_at": "2026-08-30T16:22:09+00:00" } }, { "txn_id": "BB-FRD-0417", "assessment_id": "ra_68f1863b0c8e45c4", "occurred_at": "2026-09-03T15:29:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_5d5ee2e0ca" }, "output": { "decision": "pay", "score": 0.158, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-03, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.576507+00:00", "fingerprint": "11eeb1cd888eb94be319cf3f" }, "review": null, "ledger": { "seq": 417, "hash": "f2bbf83c838b7e871058e96e2fa6c09923e05942a92b7826d815801c8e30c408", "prev_hash": "622649255d106c1eb3044b1924e8c69ebc5e1dad04906ab5bac376f219e8f847", "recorded_at": "2026-09-03T15:32:57+00:00" } }, { "txn_id": "BB-FRD-0327", "assessment_id": "ra_eda32a118b574a0f", "occurred_at": "2026-09-05T01:07:06+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_a06e65dbc5" }, "output": { "decision": "pay", "score": 0.134, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.344638+00:00", "fingerprint": "62381f203a2caa17fb54e1b6" }, "review": null, "ledger": { "seq": 327, "hash": "d6c8bebc43f9f19cbb63418d184f051203e668737f0ebf88c69bd9a6f01bdd3c", "prev_hash": "ad840ae2d4e7d4824d76125ca797a014cac30ef45cdf162b473298bf859d0c65", "recorded_at": "2026-09-05T01:08:46+00:00" } }, { "txn_id": "BB-FRD-0380", "assessment_id": "ra_7fcf17367cd24f93", "occurred_at": "2026-09-05T22:42:07+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_bd12229669" }, "output": { "decision": "pay", "score": 0.192, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.468237+00:00", "fingerprint": "115f3103335f84cc0efe0fa2" }, "review": null, "ledger": { "seq": 380, "hash": "de093bb4b7c534b06567bb8c7f29872209508abef89a0e469b77b8acfa49d1d8", "prev_hash": "8dc3e087bdfdfb869f3f9e007ddd85778bd39bed888431c687237ef99f154697", "recorded_at": "2026-09-05T22:44:10+00:00" } }, { "txn_id": "BB-FRD-0434", "assessment_id": "ra_523bda2efccb490b", "occurred_at": "2026-09-06T02:23:29+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_51ee5c3ad1" }, "output": { "decision": "pay", "score": 0.332, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-06, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.614036+00:00", "fingerprint": "39cd7fcc9878339de1ec895d" }, "review": null, "ledger": { "seq": 434, "hash": "d23d575321a879d2a4da8188303625c63d9ef43e0bb396f7860cb6f43c12da1d", "prev_hash": "eb44426761585b70e770ad530f9e6929a73fe0a8b48e0de02bb4f84f9dcacd14", "recorded_at": "2026-09-06T02:26:44+00:00" } }, { "txn_id": "BB-FRD-0404", "assessment_id": "ra_b37e5ce265e74460", "occurred_at": "2026-09-08T04:43:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_0bdf6093b6" }, "output": { "decision": "pay", "score": 0.389, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.529299+00:00", "fingerprint": "6194d8f65b599b6ca05eb21e" }, "review": null, "ledger": { "seq": 404, "hash": "753e982d2981faab847c5a79ea83482df7640c075f6806d47e3960035160ad11", "prev_hash": "2874bcd3ff65604c2e1f0995e5382e001a8a0e1bf07b3d7504d67c4299000f7d", "recorded_at": "2026-09-08T04:45:52+00:00" } }, { "txn_id": "BB-FRD-0338", "assessment_id": "ra_082e0736452f4a3a", "occurred_at": "2026-09-08T21:23:41+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_8c47eb7836" }, "output": { "decision": "pay", "score": 0.214, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-08, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.379120+00:00", "fingerprint": "ea8d3bb31b2054e4dca597d3" }, "review": null, "ledger": { "seq": 338, "hash": "e7676d1afa49a117d740ffcee9e3fe92cc036de71378f631023e91bebc8b1dd8", "prev_hash": "d3e62e28e7be014a42606bd1203491d514029e9d726953d89eea926eed60c524", "recorded_at": "2026-09-08T21:24:53+00:00" } }, { "txn_id": "BB-FRD-0423", "assessment_id": "ra_5fd513d467fb43b4", "occurred_at": "2026-09-10T16:23:40+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_754b407854" }, "output": { "decision": "pay", "score": 0.123, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.587640+00:00", "fingerprint": "4a0aada6019ba71bde70f8da" }, "review": null, "ledger": { "seq": 423, "hash": "7d097b1004e79cfdfa7b015b2723bf5cca6d30dd654800600fb59931ca0eba53", "prev_hash": "c61b5b19865c41f5369c8c0c129822fd5d44d7653534cb8fc8a436a0b174af09", "recorded_at": "2026-09-10T16:24:47+00:00" } }, { "txn_id": "BB-FRD-0364", "assessment_id": "ra_226e14cb48644c4d", "occurred_at": "2026-09-10T16:33:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_2fb37e0f0a" }, "output": { "decision": "pay", "score": 0.265, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.435961+00:00", "fingerprint": "178cac8bf0b347bebb627a1b" }, "review": null, "ledger": { "seq": 364, "hash": "1e2b17db41528b0c93da22446e3989bad1d982fee9ed17bcb72c4d597b381edb", "prev_hash": "986e6c624082bfc56acd9350116efce29abb7de67dc0682c29404aeb0a2a8782", "recorded_at": "2026-09-10T16:36:49+00:00" } }, { "txn_id": "BB-FRD-0321", "assessment_id": "ra_218d06f7d3944ecd", "occurred_at": "2026-09-12T20:59:53+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f6c4d98c79" }, "output": { "decision": "pay", "score": 0.093, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-12, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.326609+00:00", "fingerprint": "dcba1e7e5db5428c6511d0e9" }, "review": null, "ledger": { "seq": 321, "hash": "f6181ba6e6811bea0ddd8d914fc0d73700d17a9a900737469e9963505b218284", "prev_hash": "f1da339dbaf4cf1db70e52e97d15a36d9aec700e2eb787650a93b202442758c2", "recorded_at": "2026-09-12T21:01:35+00:00" } }, { "txn_id": "BB-FRD-0439", "assessment_id": "ra_f4658655b1874e6a", "occurred_at": "2026-09-14T07:21:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_c183a9d2c6" }, "output": { "decision": "pay", "score": 0.67, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.626147+00:00", "fingerprint": "e2cf104558028e0d52cedc58" }, "review": null, "ledger": { "seq": 439, "hash": "9bfc28d3a5502d0222f08c07760544b0a40feda1a94c4cf74847ad189ccf2145", "prev_hash": "93c24ff920cc36260841fb70114fedf956f3a5a73187f12a00923abb236a0453", "recorded_at": "2026-09-14T07:24:19+00:00" } }, { "txn_id": "BB-FRD-0407", "assessment_id": "ra_c1de919eb3ad411b", "occurred_at": "2026-09-14T07:37:32+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_0d308333c5" }, "output": { "decision": "pay", "score": 0.391, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.536798+00:00", "fingerprint": "caa107e82f85408992713db4" }, "review": null, "ledger": { "seq": 407, "hash": "b10510a263885f3306425d8809aa7d4895c0b45a1dd9f5997759d01a42098a36", "prev_hash": "07740fb2083df4f9ea460427abc70bff0fadc7e75885a9e48a6089fe914894b0", "recorded_at": "2026-09-14T07:38:05+00:00" } }, { "txn_id": "BB-FRD-0374", "assessment_id": "ra_fb752c9dd37e4d57", "occurred_at": "2026-09-15T09:40:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_848733bd06" }, "output": { "decision": "flag", "score": 0.98, "reason_codes": [ "F07" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "adjuster", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-15, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'flag' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 4 rule(s) in force on 2026-09-15 and recorded 3 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (high, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.457920+00:00", "fingerprint": "76e22cf0408ad694d1ffde91" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 16.7, "reviewer_role": "siu_investigator", "resolved_at": "2026-09-16T02:22:49.513375+00:00" }, "ledger": { "seq": 374, "hash": "d7a469c16dde736385d245711f65250030fafa36a70d58b3575b79046f4cf9a0", "prev_hash": "98afa7f0489ca3870049b7d8f63237fb128150a1ea72fffc8d8197b235dea8b6", "recorded_at": "2026-09-15T09:42:28+00:00" } }, { "txn_id": "BB-FRD-0428", "assessment_id": "ra_6a0f7fd34f2146ff", "occurred_at": "2026-09-16T04:34:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f80d461597" }, "output": { "decision": "pay", "score": 0.165, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-16, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.598692+00:00", "fingerprint": "739e30ed9722133b12de4fce" }, "review": null, "ledger": { "seq": 428, "hash": "72e79e6bafeffa7a5201aa8f8d06ed8d2dc5d4fee3e3c6c5d437688bc6141465", "prev_hash": "880c2682029abad40f4ab861858a0503582660970733d06f7f671753cf645226", "recorded_at": "2026-09-16T04:35:57+00:00" } }, { "txn_id": "BB-FRD-0352", "assessment_id": "ra_d55629638041453a", "occurred_at": "2026-09-17T10:41:10+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_8bfb6e442a" }, "output": { "decision": "pay", "score": 0.319, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.410278+00:00", "fingerprint": "70d3268d0dcee910a88f4e5e" }, "review": null, "ledger": { "seq": 352, "hash": "55af45b3a934c5ea6ddab360371ca4ab8d0e03e37af525a0c2e9d085c0eb6aae", "prev_hash": "791c3272fbd8e9177c66cf71a90dc6629870709673ae0ee2287c2615707d116b", "recorded_at": "2026-09-17T10:42:24+00:00" } }, { "txn_id": "BB-FRD-0381", "assessment_id": "ra_9e171ad3d5bc4bda", "occurred_at": "2026-09-17T12:39:50+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_ec7410a6b2" }, "output": { "decision": "pay", "score": 0.299, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.470790+00:00", "fingerprint": "2feb7c779377606927d5d4e3" }, "review": null, "ledger": { "seq": 381, "hash": "8549c1830e1a009d305c719a7ed8344d879ecaf4a89cae820c6f5cf6288d514e", "prev_hash": "de093bb4b7c534b06567bb8c7f29872209508abef89a0e469b77b8acfa49d1d8", "recorded_at": "2026-09-17T12:42:26+00:00" } }, { "txn_id": "BB-FRD-0414", "assessment_id": "ra_00f94c24a0954f72", "occurred_at": "2026-09-17T21:28:14+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1e42477628" }, "output": { "decision": "pay", "score": 0.085, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.567493+00:00", "fingerprint": "c0cbc7392078c4575e896f99" }, "review": null, "ledger": { "seq": 414, "hash": "46f4ae19ff5883b32c1675e3d483ea4fc069f70687462e592858000a71708bb7", "prev_hash": "40cb1bf69fc9f2ac0d89814616ab6a1dfcfe38ca628cc3d4a8f192f1a945eae8", "recorded_at": "2026-09-17T21:31:04+00:00" } }, { "txn_id": "BB-FRD-0336", "assessment_id": "ra_f50ef49f773944bb", "occurred_at": "2026-09-18T20:33:53+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_451700826c" }, "output": { "decision": "pay", "score": 0.131, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-18, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.375431+00:00", "fingerprint": "fa4788ba4c6afa34006615c5" }, "review": null, "ledger": { "seq": 336, "hash": "ae9248609b01425f2e3ce50dd94db49ef6008259c282069e34565340dc31ed76", "prev_hash": "41776227e405b64d9df6ddfb5d271582169e13b8f69a9c7f406ac2f2e8dbd483", "recorded_at": "2026-09-18T20:37:10+00:00" } }, { "txn_id": "BB-FRD-0435", "assessment_id": "ra_72f7beb119894c88", "occurred_at": "2026-09-19T04:55:55+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_7766ebd30c" }, "output": { "decision": "pay", "score": 0.261, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.617384+00:00", "fingerprint": "b7497f4933493ff275e73dca" }, "review": null, "ledger": { "seq": 435, "hash": "53af4141c885cf1285c2eb30eeef1e75a0695b2ce01f258dc7eae4e4d103a6b2", "prev_hash": "d23d575321a879d2a4da8188303625c63d9ef43e0bb396f7860cb6f43c12da1d", "recorded_at": "2026-09-19T04:57:08+00:00" } }, { "txn_id": "BB-FRD-0389", "assessment_id": "ra_4ad68f472ef448e1", "occurred_at": "2026-09-19T06:01:26+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_bdd1769c2a" }, "output": { "decision": "pay", "score": 0.213, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.496263+00:00", "fingerprint": "f266c560b8df2483dddddc16" }, "review": null, "ledger": { "seq": 389, "hash": "9bf89fe4d8d5278874b90fabd34c064f56f972860f6c8ae643ef538a8511b8ab", "prev_hash": "601e3577323438ef2ab424cfd245501a35d0329475902bb286f8b6451614529a", "recorded_at": "2026-09-19T06:03:18+00:00" } }, { "txn_id": "BB-FRD-0350", "assessment_id": "ra_d292b3306af241f1", "occurred_at": "2026-09-19T19:19:53+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_fa4256e9fd" }, "output": { "decision": "pay", "score": 0.143, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.403834+00:00", "fingerprint": "01c659de21c67231b1f4dc5d" }, "review": null, "ledger": { "seq": 350, "hash": "add1a606c6f45f64d2be7411806e8cb2aa67164173bef2b0dae9720536c17953", "prev_hash": "42eeaa3f60755ea2b00c0f68f7da613b8ffa6845164a84c4572beafbebc95c87", "recorded_at": "2026-09-19T19:23:31+00:00" } }, { "txn_id": "BB-FRD-0322", "assessment_id": "ra_29a29a8c83e04be4", "occurred_at": "2026-09-19T21:06:51+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_7cb39b1ac0" }, "output": { "decision": "pay", "score": 0.181, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-19, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.330220+00:00", "fingerprint": "04347d32d63958caa6fa04af" }, "review": null, "ledger": { "seq": 322, "hash": "d862db71e02d6807eacd6f1d149accf191afd2b0ab9364e536360f48eb7bbb2a", "prev_hash": "f6181ba6e6811bea0ddd8d914fc0d73700d17a9a900737469e9963505b218284", "recorded_at": "2026-09-19T21:10:51+00:00" } }, { "txn_id": "BB-FRD-0385", "assessment_id": "ra_93e94a68f8af453f", "occurred_at": "2026-09-20T19:26:59+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_e25fef956d" }, "output": { "decision": "pay", "score": 0.308, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-20, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.485551+00:00", "fingerprint": "d797691ecd7bdf44e2b5c6dd" }, "review": null, "ledger": { "seq": 385, "hash": "b159c31eba1fb665f4c6414da55ec4ec2f2120308e98bf267582e55a50f9926c", "prev_hash": "aed3eb50c1bc245659791b94f096a17fd51acbcf51dea344458917ab4f579678", "recorded_at": "2026-09-20T19:29:54+00:00" } }, { "txn_id": "BB-FRD-0330", "assessment_id": "ra_aea1fd591dcc4f91", "occurred_at": "2026-09-21T18:34:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_a799a907c5" }, "output": { "decision": "pay", "score": 0.121, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.351053+00:00", "fingerprint": "d7599c0c9cce045ca2f96639" }, "review": null, "ledger": { "seq": 330, "hash": "52f318368018ecc75b48ccf3912880a0aea3880b35535a9a1db2574765afeff4", "prev_hash": "516a1a4b876a558dacfd6359089e62a8b028b99c816024c20ece6b0e69bdbafe", "recorded_at": "2026-09-21T18:37:57+00:00" } }, { "txn_id": "BB-FRD-0376", "assessment_id": "ra_df3592e50873432d", "occurred_at": "2026-09-24T14:05:59+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1e938ffbba" }, "output": { "decision": "pay", "score": 0.35, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-24, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.461523+00:00", "fingerprint": "60ba337146c0bdedd0a90fc1" }, "review": null, "ledger": { "seq": 376, "hash": "3627327df442825095448eb94cc93b6c2887416165b2b8483df3257051c00f8e", "prev_hash": "7cebccc2072c5cf04bd2c36b859eba4f2b110ec9757e58c0f6d0da9062e4f24b", "recorded_at": "2026-09-24T14:08:25+00:00" } }, { "txn_id": "BB-FRD-0436", "assessment_id": "ra_436d8dee231f4cc2", "occurred_at": "2026-09-24T16:36:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_03b3444c9c" }, "output": { "decision": "pay", "score": 0.327, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-24, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.619319+00:00", "fingerprint": "cbd2a6c4578d1b858353294c" }, "review": null, "ledger": { "seq": 436, "hash": "6d8fbd35c41d4955b4d2e67045d8bda637a3d3b3126ccdd2cd0a12cf5e5e76ba", "prev_hash": "53af4141c885cf1285c2eb30eeef1e75a0695b2ce01f258dc7eae4e4d103a6b2", "recorded_at": "2026-09-24T16:39:13+00:00" } }, { "txn_id": "BB-FRD-0429", "assessment_id": "ra_a943e5c243574da9", "occurred_at": "2026-09-26T21:24:57+00:00", "state": "CO", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_310baa4c62" }, "output": { "decision": "pay", "score": 0.062, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-26, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.600456+00:00", "fingerprint": "66bfbcba0740aa7cf7d171a0" }, "review": null, "ledger": { "seq": 429, "hash": "2b537b61d5b726629f9588098661e15b66d2080edbc77a17383880a40f552377", "prev_hash": "72e79e6bafeffa7a5201aa8f8d06ed8d2dc5d4fee3e3c6c5d437688bc6141465", "recorded_at": "2026-09-26T21:28:17+00:00" } }, { "txn_id": "BB-FRD-0355", "assessment_id": "ra_60223024d21745a0", "occurred_at": "2026-09-29T07:17:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_85cf49d34d" }, "output": { "decision": "pay", "score": 0.453, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.415894+00:00", "fingerprint": "7d974541d3870236a3e9c650" }, "review": null, "ledger": { "seq": 355, "hash": "682c87ed574c5fc87624b9b295906b5b7b9893f60cc8e4d863c430c36a6a8654", "prev_hash": "c196025f764507165b22ce8f89e5b50cae13799ba073c7cd7bec57583a992ae6", "recorded_at": "2026-09-29T07:20:58+00:00" } }, { "txn_id": "BB-FRD-0408", "assessment_id": "ra_20756c38315d4008", "occurred_at": "2026-09-29T23:45:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_d75c152932" }, "output": { "decision": "pay", "score": 0.093, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.543369+00:00", "fingerprint": "e4e51c27c3e2caea20d6a351" }, "review": null, "ledger": { "seq": 408, "hash": "770bc6f4deb31fe6d4580cc4cbae96187016221ef6ee17425044fd68f12ad770", "prev_hash": "b10510a263885f3306425d8809aa7d4895c0b45a1dd9f5997759d01a42098a36", "recorded_at": "2026-09-29T23:46:52+00:00" } }, { "txn_id": "BB-FRD-0419", "assessment_id": "ra_1dad19355bde4f4b", "occurred_at": "2026-09-30T14:54:13+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_ae58d48636" }, "output": { "decision": "pay", "score": 0.399, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.580238+00:00", "fingerprint": "c444e50611f35165f4daa3d9" }, "review": null, "ledger": { "seq": 419, "hash": "99742716f189d1ad79905283e1b40cf994cc24eefcedc94acac1e0495202765e", "prev_hash": "45017ac2295b740b7ea442232d64691c93dd996a87c3ca73848dd446f6d27cb7", "recorded_at": "2026-09-30T14:55:57+00:00" } }, { "txn_id": "BB-FRD-0418", "assessment_id": "ra_d52d61610de94cf9", "occurred_at": "2026-10-03T20:52:53+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_eddf897baf" }, "output": { "decision": "pay", "score": 0.083, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-03, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.577999+00:00", "fingerprint": "963caf4280787513fc57c57d" }, "review": null, "ledger": { "seq": 418, "hash": "45017ac2295b740b7ea442232d64691c93dd996a87c3ca73848dd446f6d27cb7", "prev_hash": "f2bbf83c838b7e871058e96e2fa6c09923e05942a92b7826d815801c8e30c408", "recorded_at": "2026-10-03T20:56:03+00:00" } }, { "txn_id": "BB-FRD-0431", "assessment_id": "ra_496b85cf0b614c79", "occurred_at": "2026-10-04T06:29:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_261521fc33" }, "output": { "decision": "pay", "score": 0.227, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-04, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.603857+00:00", "fingerprint": "de7e5fa8a1731fe65901e8cc" }, "review": null, "ledger": { "seq": 431, "hash": "72baa6f66e01a70281cdb88400b72025984cfa028d18534ec81c203aaf18eeda", "prev_hash": "26f5c7353fa09839ff7a872b0e13e82b963765b4935a6e3337fb367ded6335e8", "recorded_at": "2026-10-04T06:33:45+00:00" } }, { "txn_id": "BB-FRD-0325", "assessment_id": "ra_bcc227e727de4046", "occurred_at": "2026-10-04T07:39:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f2322083e8" }, "output": { "decision": "pay", "score": 0.095, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-04, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.337321+00:00", "fingerprint": "c133b07cefda7c9c79e79830" }, "review": null, "ledger": { "seq": 325, "hash": "60dc85d797fb6c6f0774de221f09d50ae2d3f3db42ff8bae21bec0ba3509ea83", "prev_hash": "3ac0db999b797e561ea716a021a8f855d08cc0736b95a4399cb74f53f2a0c389", "recorded_at": "2026-10-04T07:42:03+00:00" } }, { "txn_id": "BB-FRD-0332", "assessment_id": "ra_7b7061d28fa14806", "occurred_at": "2026-10-04T11:04:03+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f5bae936f8" }, "output": { "decision": "pay", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-04, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.357350+00:00", "fingerprint": "db08efb860652ef0a5681464" }, "review": null, "ledger": { "seq": 332, "hash": "21aecd1322daa424ec6d8638cc207d71470a920b34964c9251c59cf0762f0792", "prev_hash": "de131fd8cf1112fe70ba718f2d45e2a745095b7cabbaf4c1eb338be376a6bbda", "recorded_at": "2026-10-04T11:05:04+00:00" } }, { "txn_id": "BB-FRD-0412", "assessment_id": "ra_fc98b6c910d54e3e", "occurred_at": "2026-10-05T12:17:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_321771d019" }, "output": { "decision": "pay", "score": 0.467, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.562037+00:00", "fingerprint": "987782bafa8498294215a0a8" }, "review": null, "ledger": { "seq": 412, "hash": "e9143fd6976a036c090b371c67a7724ba55cfbd7687855149a40b3f93b2efe4f", "prev_hash": "f747d9c78b7bba09560edb7d605ad78e6dfb8fb8887720f3762f0cb2d9b73cc2", "recorded_at": "2026-10-05T12:18:24+00:00" } }, { "txn_id": "BB-FRD-0432", "assessment_id": "ra_4157c05498e74ee8", "occurred_at": "2026-10-05T13:16:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_7b61c5c129" }, "output": { "decision": "pay", "score": 0.031, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.607421+00:00", "fingerprint": "375cbf8197c4d68bcd25d355" }, "review": null, "ledger": { "seq": 432, "hash": "7179f61f54dcccdc968958b353db6080c0af7f7843453e40425d57d01b93e2cb", "prev_hash": "72baa6f66e01a70281cdb88400b72025984cfa028d18534ec81c203aaf18eeda", "recorded_at": "2026-10-05T13:17:49+00:00" } }, { "txn_id": "BB-FRD-0424", "assessment_id": "ra_590d2fb9cdf24c1f", "occurred_at": "2026-10-05T20:01:17+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_e911385944" }, "output": { "decision": "pay", "score": 0.291, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-05, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.591152+00:00", "fingerprint": "4f13300e19b95f284cee3827" }, "review": null, "ledger": { "seq": 424, "hash": "4f1c5a3de37738a086a59b2bf7af80686b63d71f3dec97a1c344d399b5b435ec", "prev_hash": "7d097b1004e79cfdfa7b015b2723bf5cca6d30dd654800600fb59931ca0eba53", "recorded_at": "2026-10-05T20:04:45+00:00" } }, { "txn_id": "BB-FRD-0356", "assessment_id": "ra_140b510d34ea4355", "occurred_at": "2026-10-06T05:35:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_1dcd621162" }, "output": { "decision": "pay", "score": 0.213, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-06, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.417726+00:00", "fingerprint": "5d689e4b56149a662766f1a9" }, "review": null, "ledger": { "seq": 356, "hash": "1d2e06cfb11af029f91a3c2b769f935e56320a9243317b0caf29837ebcda4b83", "prev_hash": "682c87ed574c5fc87624b9b295906b5b7b9893f60cc8e4d863c430c36a6a8654", "recorded_at": "2026-10-06T05:36:17+00:00" } }, { "txn_id": "BB-FRD-0328", "assessment_id": "ra_9d4a169eca3043ae", "occurred_at": "2026-10-06T07:28:08+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_ba563e6b13" }, "output": { "decision": "pay", "score": 0.459, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-06, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.347420+00:00", "fingerprint": "5637431c2c298593b560a437" }, "review": null, "ledger": { "seq": 328, "hash": "2ad7b6b38ad5eb56bd9970f4d6d74a9dc10c9970652949b6975d86c601d5a272", "prev_hash": "d6c8bebc43f9f19cbb63418d184f051203e668737f0ebf88c69bd9a6f01bdd3c", "recorded_at": "2026-10-06T07:29:33+00:00" } }, { "txn_id": "BB-FRD-0384", "assessment_id": "ra_5fa165cca2eb4e73", "occurred_at": "2026-10-07T07:10:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_2b06efe3ec" }, "output": { "decision": "pay", "score": 0.215, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.482984+00:00", "fingerprint": "8f68469064b8e920ac85a482" }, "review": null, "ledger": { "seq": 384, "hash": "aed3eb50c1bc245659791b94f096a17fd51acbcf51dea344458917ab4f579678", "prev_hash": "5219e3780ac8a440f7bc4a180f99f9c9a41d98df5b3467d5d5abb073cd9a8bca", "recorded_at": "2026-10-07T07:13:30+00:00" } }, { "txn_id": "BB-FRD-0383", "assessment_id": "ra_88f472d03e174078", "occurred_at": "2026-10-07T13:26:07+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_7a472190a2" }, "output": { "decision": "pay", "score": 0.356, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.480574+00:00", "fingerprint": "b6414d822a37670672604630" }, "review": null, "ledger": { "seq": 383, "hash": "5219e3780ac8a440f7bc4a180f99f9c9a41d98df5b3467d5d5abb073cd9a8bca", "prev_hash": "cdb4d326cfc39159e9651d5bb076d002128cebf4dfc820565f7aca36f6bbb04d", "recorded_at": "2026-10-07T13:27:19+00:00" } }, { "txn_id": "BB-FRD-0369", "assessment_id": "ra_0294c634054f4b95", "occurred_at": "2026-10-09T01:08:38+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_38ed9df87a" }, "output": { "decision": "pay", "score": 0.193, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.447392+00:00", "fingerprint": "8dece06c1e1531c219b14d26" }, "review": null, "ledger": { "seq": 369, "hash": "867c469328cf326604c4cb7b9b2684660dc052a409ff1101db260c301b483ae7", "prev_hash": "29391975e8d05533a0360aeb54159358bca8bdd39f4a517912c5a00dcb66eaad", "recorded_at": "2026-10-09T01:09:43+00:00" } }, { "txn_id": "BB-FRD-0437", "assessment_id": "ra_a49e17d862be47db", "occurred_at": "2026-10-10T06:11:20+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_50e09b9d4c" }, "output": { "decision": "pay", "score": 0.158, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-10, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.620931+00:00", "fingerprint": "d46a0e43b1093036bdea67cf" }, "review": null, "ledger": { "seq": 437, "hash": "2cca4f2db462640888a9691be667d4a72e98c19c53634ae7070f0f903692da84", "prev_hash": "6d8fbd35c41d4955b4d2e67045d8bda637a3d3b3126ccdd2cd0a12cf5e5e76ba", "recorded_at": "2026-10-10T06:14:23+00:00" } }, { "txn_id": "BB-FRD-0342", "assessment_id": "ra_9cc33ff5a9fd49cb", "occurred_at": "2026-10-11T13:08:15+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_8d81f83eed" }, "output": { "decision": "pay", "score": 0.086, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-11, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.387074+00:00", "fingerprint": "dc84ab06d1276568cdcebca4" }, "review": null, "ledger": { "seq": 342, "hash": "89667068b1213bc7a60d4f97264050f2a40087e0f754aa392ce5c5373a27b5df", "prev_hash": "4b7f6553311301da25f9ac5de5c0cc34ed6ceb5ea67b55b198c11991e7ff991d", "recorded_at": "2026-10-11T13:09:37+00:00" } }, { "txn_id": "BB-FRD-0416", "assessment_id": "ra_dc6789a798de4ca9", "occurred_at": "2026-10-12T00:34:49+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_cce531c540" }, "output": { "decision": "pay", "score": 0.042, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.574801+00:00", "fingerprint": "00ead4358b1abc291f5df6ed" }, "review": null, "ledger": { "seq": 416, "hash": "622649255d106c1eb3044b1924e8c69ebc5e1dad04906ab5bac376f219e8f847", "prev_hash": "0aae165d74f6cffd258f1d9e8c5da53447f4305dad999709027702af4608db2b", "recorded_at": "2026-10-12T00:38:12+00:00" } }, { "txn_id": "BB-FRD-0345", "assessment_id": "ra_5049c2f8728c4045", "occurred_at": "2026-10-12T03:47:50+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_f17c5600cc" }, "output": { "decision": "pay", "score": 0.206, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.394124+00:00", "fingerprint": "cb1223bdcecd3b9a26b7bef0" }, "review": null, "ledger": { "seq": 345, "hash": "0510035612beed985551e6b8fd7dc44177dfd384fb47a0cdf8149b366dfb7b7e", "prev_hash": "d7c0e3f4abbcc3a42f1f9fa18ed4cae6f252f65005b95ae6f7cadac06924474b", "recorded_at": "2026-10-12T03:49:36+00:00" } }, { "txn_id": "BB-FRD-0349", "assessment_id": "ra_430179c529354992", "occurred_at": "2026-10-12T13:23:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_aea3e95c9d" }, "output": { "decision": "pay", "score": 0.25, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.401701+00:00", "fingerprint": "30f6f87733ecc22bac7a2cee" }, "review": null, "ledger": { "seq": 349, "hash": "42eeaa3f60755ea2b00c0f68f7da613b8ffa6845164a84c4572beafbebc95c87", "prev_hash": "9dc846aceee2458f567486d67eeadd813f8335416eae800f2ce50d73bb3d5ca9", "recorded_at": "2026-10-12T13:25:27+00:00" } }, { "txn_id": "BB-FRD-0395", "assessment_id": "ra_3ad87029578940f9", "occurred_at": "2026-10-13T08:40:07+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_9747a23a45" }, "output": { "decision": "pay", "score": 0.384, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-13, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.509933+00:00", "fingerprint": "d4f0a42ee80989028b38aff0" }, "review": null, "ledger": { "seq": 395, "hash": "25292fef1df8378139704600107f01259346b3f9a52f48425251e3ecc6fa25a3", "prev_hash": "3a80c1fc7cfee26dca115b62c082db1d61ac178010afb04f52f80963396e5682", "recorded_at": "2026-10-13T08:41:52+00:00" } }, { "txn_id": "BB-FRD-0354", "assessment_id": "ra_1d6c8a8b9b5e4bf6", "occurred_at": "2026-10-13T09:48:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_0f457b7663" }, "output": { "decision": "pay", "score": 0.492, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-13, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.414172+00:00", "fingerprint": "8233f5841c701a6f4514c4a8" }, "review": null, "ledger": { "seq": 354, "hash": "c196025f764507165b22ce8f89e5b50cae13799ba073c7cd7bec57583a992ae6", "prev_hash": "c7184c789a780b4685e12020784f54ebc4bef9367ecbe951750095a79164556b", "recorded_at": "2026-10-13T09:50:43+00:00" } }, { "txn_id": "BB-FRD-0415", "assessment_id": "ra_6e31feeb15334c06", "occurred_at": "2026-10-14T15:49:06+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_24e438f48b" }, "output": { "decision": "pay", "score": 0.411, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.570545+00:00", "fingerprint": "09d21d77f8cc8d5a140eea25" }, "review": null, "ledger": { "seq": 415, "hash": "0aae165d74f6cffd258f1d9e8c5da53447f4305dad999709027702af4608db2b", "prev_hash": "46f4ae19ff5883b32c1675e3d483ea4fc069f70687462e592858000a71708bb7", "recorded_at": "2026-10-14T15:50:35+00:00" } }, { "txn_id": "BB-FRD-0331", "assessment_id": "ra_18572a1077044f9e", "occurred_at": "2026-10-14T16:28:33+00:00", "state": "TX", "line_of_business": "home", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_4937121b23" }, "output": { "decision": "pay", "score": 0.347, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX home fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.353450+00:00", "fingerprint": "142bdf1c3861bee361de2e2f" }, "review": null, "ledger": { "seq": 331, "hash": "de131fd8cf1112fe70ba718f2d45e2a745095b7cabbaf4c1eb338be376a6bbda", "prev_hash": "52f318368018ecc75b48ccf3912880a0aea3880b35535a9a1db2574765afeff4", "recorded_at": "2026-10-14T16:30:06+00:00" } }, { "txn_id": "BB-FRD-0372", "assessment_id": "ra_c18964e47ff24c5c", "occurred_at": "2026-10-14T23:37:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_3e40392322" }, "output": { "decision": "pay", "score": 0.233, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.452477+00:00", "fingerprint": "0c2aa8fbde124464e415ce4f" }, "review": null, "ledger": { "seq": 372, "hash": "94328e8be38b94a1a0407274a95d233f2c83fdaf64cf329b96ce017266fa608b", "prev_hash": "519b6789c4a9e688582e7d32ffcab3daadff1098ac1e5ffeba2137d0f01b58a7", "recorded_at": "2026-10-14T23:40:06+00:00" } }, { "txn_id": "BB-FRD-0400", "assessment_id": "ra_d9fbb7d2fd5345b7", "occurred_at": "2026-10-16T07:05:56+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_2e4f6cf200" }, "output": { "decision": "pay", "score": 0.227, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-16, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.519637+00:00", "fingerprint": "5be029958d703ef85613ab6f" }, "review": null, "ledger": { "seq": 400, "hash": "7f0e994eeaf2dd4954e16d1ce169c25afbd365cca64d9073c925c5497763564d", "prev_hash": "6ca2e53d1e5cbd34f5d4d3c2fc78199746c37a90896f94bf9430dae8caa847b8", "recorded_at": "2026-10-16T07:06:49+00:00" } }, { "txn_id": "BB-FRD-0440", "assessment_id": "ra_3de8d11de81e4ee2", "occurred_at": "2026-10-17T21:44:06+00:00", "state": "TX", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_8301f02626" }, "output": { "decision": "pay", "score": 0.186, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-17, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a TX auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 4 rule(s) in force on 2026-10-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.628404+00:00", "fingerprint": "0e6eacfd52618c5209c17549" }, "review": null, "ledger": { "seq": 440, "hash": "352007dcd2195f4e41be7ff432adf61a68624463c86d7adbbe21fe188a90f169", "prev_hash": "9bfc28d3a5502d0222f08c07760544b0a40feda1a94c4cf74847ad189ccf2145", "recorded_at": "2026-10-17T21:44:41+00:00" } }, { "txn_id": "TX-CLM-0417", "assessment_id": "ra_3e5ace53679a49aa", "occurred_at": "2026-10-21T09:00:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_showcase_0417" }, "output": { "decision": "deny", "score": 0.93, "reason_codes": [ "F07", "F12" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": "claims@bluebonnet.example" }, "carrier_note": "Rear-end collision claim; vendor fraud score 0.93 on social-media and location signals; denial letter queued automatically.", "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'deny' in a TX auto claims transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-10-21 and recorded 4 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (critical, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "critical", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:25.350669+00:00", "fingerprint": "422af91fd8483ebae2883e3b" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 3.2, "reviewer_role": "adjuster", "resolved_at": "2026-10-21T12:14:27.984090+00:00" }, "ledger": { "seq": 1121, "hash": "dba2ecd14e3ab92057b32340d5c7a1dcb8a74ac3c1f828915a3a2db979d42971", "prev_hash": "078f70f823f4608edd011aeed4df5100cecd0c7d0762b43639ae31c7e4ecf30d", "recorded_at": "2026-10-21T09:00:50+00:00" } }, { "txn_id": "TX-CLM-0417-LIVE-851", "assessment_id": "ra_1414e06b06fb4321", "occurred_at": "2026-10-21T09:00:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "claims", "program_area": "Claims/Adjudication", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_showcase_0417" }, "output": { "decision": "deny", "score": 0.93, "reason_codes": [ "F07", "F12" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": "claims@bluebonnet.example" }, "carrier_note": "Rear-end collision claim; vendor fraud score 0.93 on social-media and location signals; denial letter queued automatically.", "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'deny' in a TX auto claims transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 7 rule(s) in force on 2026-10-21 and recorded 4 finding(s); the most serious is 'Algorithmic denial, reduction or fraud hold with no human review' (critical, L2).", "findings": [ { "rule_id": "tx.542.algorithmic-denial-without-human-review", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "critical", "title": "Algorithmic denial, reduction or fraud hold with no human review", "required_action": "Route to an adjuster for review before the outcome is communicated; record the reviewer and outcome.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-055", "label": "Tex. Ins. Code \u00a7 542.055 \u2014 receipt of notice of claim (15-day acknowledgement)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.542.rejection-without-reasons", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Claim rejected or reduced without stated reasons", "required_action": "Hold the rejection until the reasons are stated in the claimant notice and recorded; confirm the 15-business-day clock.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-542-056", "label": "Tex. Ins. Code \u00a7 542.056 \u2014 notice of acceptance or rejection of claim; reasons for rejection", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true }, { "provision_id": "tx-ins-542-003", "label": "Tex. Ins. Code \u00a7 542.003 \u2014 unfair claim settlement practices prohibited", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.542", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:26:26.589955+00:00", "fingerprint": "422af91fd8483ebae2883e3b" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 0.0, "reviewer_role": "compliance", "resolved_at": "2026-10-30T11:26:36.167701+00:00" }, "ledger": { "seq": 1372, "hash": "0a2bfdac9c8f0e74640b87e0e2c969a1015001d84c76beba7f000b7872357939", "prev_hash": "f288edc347d82bbc6569dfac8ff150d4d09f0d602fdfb2f43a466120011d1d98", "recorded_at": "2026-10-30T11:26:26.601484+00:00" } }, { "txn_id": "BB-FRD-0337", "assessment_id": "ra_3afce55dc56743ae", "occurred_at": "2026-10-21T19:35:03+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_9fd8072bbf" }, "output": { "decision": "pay", "score": 0.351, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-21, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.377285+00:00", "fingerprint": "276e0ec0e0e1e0b8bda9a16d" }, "review": null, "ledger": { "seq": 337, "hash": "d3e62e28e7be014a42606bd1203491d514029e9d726953d89eea926eed60c524", "prev_hash": "ae9248609b01425f2e3ce50dd94db49ef6008259c282069e34565340dc31ed76", "recorded_at": "2026-10-21T19:36:05+00:00" } }, { "txn_id": "BB-FRD-0363", "assessment_id": "ra_3f703c47127347cb", "occurred_at": "2026-10-22T18:54:58+00:00", "state": "CO", "line_of_business": "auto", "stage": "fraud", "program_area": "Fraud/Waste & Abuse", "ai_system": { "system_id": "fnol-fraud", "model_version": "3.1", "vendor": "third_party", "vendor_name": "Sentinel Analytics (fictional)", "high_risk_class": "fraud" }, "inputs": { "feature_classes": [ "claims_history", "geography", "social_media" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "vendor://sentinel/fnol" ], "consumer_token": "tok_7b2cda7b64" }, "output": { "decision": "pay", "score": 0.064, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, fnol-fraud v3.1 (a third-party system from Sentinel Analytics (fictional)) produced 'pay' in a CO auto fraud transaction. Inputs included claims history, geography, social media. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 1 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.434271+00:00", "fingerprint": "c4e03efc1e85c0ca46fb7e5d" }, "review": null, "ledger": { "seq": 363, "hash": "986e6c624082bfc56acd9350116efce29abb7de67dc0682c29404aeb0a2a8782", "prev_hash": "9479c6d8bfdc12f12d30ff79f38ae5bb5e827ae6ea3fd1c2f3f486344903e32d", "recorded_at": "2026-10-22T18:56:44+00:00" } } ] }, { "system_id": "price-opt", "items": { "1_name_version": "Renewal Price Optimizer \u2014 versions observed: 7.2 (declared 7.1)", "2_use_cases": "Recommends renewal premium within filed rating plan bounds using telematics, credit-based insurance score and territory. Stages: {'renewal': 168}; decisions in period: 168; adverse: 45; states: {'TX': 116, 'CO': 52}.", "3_model_type": "gradient-boosted ML", "3a_agentic": { "agentic": false, "objectives": null, "authorizations_without_human_approval": null }, "4_implementation_date": "2025-01-15", "5_development": { "vendor": "internal", "vendor_name": null }, "6_purchased_modifiable": { "purchased": false, "developed_for_company": true, "modifiable": true }, "7_model_risks": { "description": "Potential adverse consumer impact evidenced by findings on the model's decisions in the period.", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "severity": "medium", "layer": "L6", "count": 45 }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "title": "Adverse outcome from a high-risk AI system with no human review", "severity": "medium", "layer": "L6", "count": 35 }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "severity": "high", "layer": "L4", "count": 28 }, { "rule_id": "co.reg-10-1-1.governance-framework-amended", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "severity": "medium", "layer": "L3", "count": 26 }, { "rule_id": "tx.559.credit-adverse-effect-notice", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "severity": "high", "layer": "L2", "count": 8 } ], "adverse_decisions": 45, "dispositions": { "HOLD": 8, "PASS": 105, "FLAG": 55 } }, "8_limitations": { "carrier_documentation": "no model documentation reference recorded on the decision stream", "regbench_note": "RegBench assesses recorded facts about each decision; it does not inspect model internals or detect inputs outside the training domain." }, "9_automate_augment_support": { "declared": "automate", "observed_human_review_pct": 19.0 }, "10_testing_validation": { "testing_report_ref_on_decisions": [ "test://bluebonnet/price-opt/2026Q2" ], "regbench_continuous_assessment": "168 of 168 decisions assessed at decision time against the rules in force on each decision date." }, "10a_proxy_evaluation": [ { "system_id": "price-opt", "model_version": "7.2", "tenant_id": "bluebonnet", "state": null, "stage_mix": { "renewal": 168 }, "reference_group": "A", "groups": [ { "group": "A", "decisions": 100, "adverse": 22, "adverse_rate": 0.22 } ], "suppressed_groups": [ "D", "B", "C" ], "flags": [], "method": "Adverse-outcome rate by carrier-estimated demographic proxy group; reference group = largest group; adverse impact ratio = group rate / reference rate (flag below 0.80); marginal difference in percentage points (flag at 5.0 or more). Cells under 30 decisions are suppressed. Illustrative thresholds.", "computed_where": "carrier environment (proxy labels never leave it); only these aggregates are released" } ], "11_last_test_date": { "carrier": "see testing report reference", "regbench_last_assessment": "2026-10-27T17:36:12+00:00" }, "12_financial_statement_effect": "material", "13_legal_compliance_review": { "rules_applied": [ "co.reg-10-1-1.governance-framework-amended", "doctrine.unreviewed-adverse-high-risk", "doctrine.zip-credit-interaction-proxy", "tx.559.credit-adverse-effect-notice", "tx.tdi.consequential-decision-without-human-review" ], "provisions_cited": [ "co-reg-10-1-1-s5", "co-reg-10-1-1-s6", "naic-ai-bulletin-s3-1-ais-program", "tx-ins-559-001", "tx-ins-559-054", "tx-tdi-b-0003-26-governance", "tx-tdi-b-0003-26-human-review" ], "method": "Deterministic rule evaluation per decision; citations resolve against the corpus snapshot in force on the decision date or the assessment fails closed." }, "14_regulatory_actions": "None recorded in RegBench. Carrier to confirm any informal agreements, administrative complaints, consent orders or investigations." }, "case_files": [ { "txn_id": "BB-RNW-0295", "assessment_id": "ra_8e9381e2cb2c43ee", "occurred_at": "2026-07-02T23:41:23+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_53a5c2d943" }, "output": { "decision": "rate", "score": 0.781, "reason_codes": [], "premium_delta_pct": 15.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-05T23:41:23+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-05 with reasons stated. RegBench applied 11 rule(s) in force on 2026-07-02 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.260100+00:00", "fingerprint": "2dfce338d301804124bc096b" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 4.7, "reviewer_role": "compliance", "resolved_at": "2026-07-03T04:24:39.417530+00:00" }, "ledger": { "seq": 295, "hash": "a71e0a7dbcc4c7e6c5bb92fc99c7b70068ac74c6e976b33a11f1a89e644f75ed", "prev_hash": "4714b66d15cb1982fd5a47e7d105d102f79b63f4f3829d0f066e9ea63e16b8aa", "recorded_at": "2026-07-02T23:44:10+00:00" } }, { "txn_id": "BB-RNW-0182", "assessment_id": "ra_1139da7766d64f48", "occurred_at": "2026-07-03T02:04:40+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_69922224c0" }, "output": { "decision": "rate", "score": 0.499, "reason_codes": [], "premium_delta_pct": -9.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-03, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.969353+00:00", "fingerprint": "bc27ea669d16ae1fe3c364d3" }, "review": null, "ledger": { "seq": 182, "hash": "211e058d2055d05c3f77f226164200c17323313240015437d8462b301151699b", "prev_hash": "8cb5401e9949b82f0df66e4bf764f79bbf341b33542dede477cf4e858a2b2cbd", "recorded_at": "2026-07-03T02:06:27+00:00" } }, { "txn_id": "BB-RNW-0178", "assessment_id": "ra_461b724105d64499", "occurred_at": "2026-07-03T07:36:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4e5c0dc65c" }, "output": { "decision": "rate", "score": 0.4, "reason_codes": [], "premium_delta_pct": -5.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.960112+00:00", "fingerprint": "8fbad82adb4d7d12f08ff6be" }, "review": null, "ledger": { "seq": 178, "hash": "f342d0c62aa197ac0450f07bf534f40c83126fa01ef9587a5a1a55c09c0ae74c", "prev_hash": "9060247a6c6f86225f0bc439289749e13c684747ee84b6d7ff0599134fa5c9e8", "recorded_at": "2026-07-03T07:39:38+00:00" } }, { "txn_id": "BB-RNW-0228", "assessment_id": "ra_4f55502757474fa9", "occurred_at": "2026-07-03T12:55:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_91124e5f37" }, "output": { "decision": "rate", "score": 0.32, "reason_codes": [], "premium_delta_pct": 0.4 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.077554+00:00", "fingerprint": "80a5689c3c5fd9111c350137" }, "review": null, "ledger": { "seq": 228, "hash": "638abc268584b7c9b939af8d3778a4aaa0b631d5090e1b432cf285b869b5295a", "prev_hash": "babc8f48561ef7e4c07d19cb371febe29a506ea422726a2ca3c74a9a3a1db462", "recorded_at": "2026-07-03T12:57:21+00:00" } }, { "txn_id": "BB-RNW-0294", "assessment_id": "ra_71ddea418eb046d2", "occurred_at": "2026-07-04T02:02:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_fc082c00e1" }, "output": { "decision": "rate", "score": 0.313, "reason_codes": [], "premium_delta_pct": 23.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-10T02:02:32+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-04, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-10 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-04 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.258024+00:00", "fingerprint": "f2f64cecde52a815f0889f9b" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 26.4, "reviewer_role": "compliance", "resolved_at": "2026-07-05T04:26:48.358951+00:00" }, "ledger": { "seq": 294, "hash": "4714b66d15cb1982fd5a47e7d105d102f79b63f4f3829d0f066e9ea63e16b8aa", "prev_hash": "a6a2e7d14721c9a7262a1e228b0a4da4c928a145ccb711aae1a4cb202fe25729", "recorded_at": "2026-07-04T02:03:04+00:00" } }, { "txn_id": "BB-RNW-0288", "assessment_id": "ra_8e410cc3149d49f3", "occurred_at": "2026-07-04T02:07:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8142fa9801" }, "output": { "decision": "rate", "score": 0.365, "reason_codes": [], "premium_delta_pct": 17.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-08T02:07:53+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-04, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-08 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-04 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.243989+00:00", "fingerprint": "7c8a380643f86ffb577feed8" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 288, "hash": "3e1e79b3604aa476f64d12e336aee3589b24e5fbbfa7880eceed91b6f4ac73c8", "prev_hash": "1df44da920e402366268f37e68fef1832c7f4c920f9a31f71e92ce59746415e6", "recorded_at": "2026-07-04T02:09:22+00:00" } }, { "txn_id": "BB-RNW-0157", "assessment_id": "ra_aa0b61a7bdc34dd3", "occurred_at": "2026-07-05T11:53:49+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_67aa4a2f89" }, "output": { "decision": "rate", "score": 0.833, "reason_codes": [], "premium_delta_pct": 0.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-05, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.912276+00:00", "fingerprint": "1b37ef8f0585e45e223d0461" }, "review": null, "ledger": { "seq": 157, "hash": "28c03f462f3fe7d9f880d773c5d84743cc7f89e71b0ea7cb683793ec659dd011", "prev_hash": "2dcc88a2a3d83e16bad82cf574bfc894aadccfdf6a677b4aadf4105aa1448cc4", "recorded_at": "2026-07-05T11:56:30+00:00" } }, { "txn_id": "BB-RNW-0310", "assessment_id": "ra_c61bf492c97a4abf", "occurred_at": "2026-07-05T18:40:42+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d21140bdd9" }, "output": { "decision": "rate", "score": 0.582, "reason_codes": [], "premium_delta_pct": -13.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-05, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.297226+00:00", "fingerprint": "732ecb4f50d92e960e356883" }, "review": null, "ledger": { "seq": 310, "hash": "8bf70166591b81e1fd31772be3645d300fbdebd78427104a11c849fc0e04c09d", "prev_hash": "e520e7f243ec61ed3bdf23f7336ad0a07f3b5c697a54d79cbf5be155bddd6939", "recorded_at": "2026-07-05T18:42:33+00:00" } }, { "txn_id": "BB-RNW-0163", "assessment_id": "ra_a0ed1b2918bf4f5a", "occurred_at": "2026-07-05T21:39:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e1c8c24e82" }, "output": { "decision": "rate", "score": 0.242, "reason_codes": [], "premium_delta_pct": 19.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-10T21:39:05+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-05, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-10 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-05 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.928725+00:00", "fingerprint": "d2a9afcbc61c56858d09ea9f" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 29.0, "reviewer_role": "compliance", "resolved_at": "2026-07-07T02:39:35.826021+00:00" }, "ledger": { "seq": 163, "hash": "899968a82b0bd419dea801f8ec2cceaa084df2a43aa0c9db7aff6e00fdf42220", "prev_hash": "9d3a3f152cd88a9b6f7af0e59f7f3e9161f7646ebca77f854bc835bdfffad3db", "recorded_at": "2026-07-05T21:42:26+00:00" } }, { "txn_id": "BB-RNW-0162", "assessment_id": "ra_1f045f24a97d4fad", "occurred_at": "2026-07-06T05:52:47+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4bbf3fc9bf" }, "output": { "decision": "rate", "score": 0.314, "reason_codes": [], "premium_delta_pct": -8.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-06, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.927175+00:00", "fingerprint": "6011dd19d0afb4ca1406fe64" }, "review": null, "ledger": { "seq": 162, "hash": "9d3a3f152cd88a9b6f7af0e59f7f3e9161f7646ebca77f854bc835bdfffad3db", "prev_hash": "8c3db1a3515723b910fabce35dc212065ea8ef52a8c3b667218b0a9cab68bbbb", "recorded_at": "2026-07-06T05:55:27+00:00" } }, { "txn_id": "BB-RNW-0312", "assessment_id": "ra_6681ce1197a14788", "occurred_at": "2026-07-07T05:22:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9010983c42" }, "output": { "decision": "rate", "score": 0.61, "reason_codes": [], "premium_delta_pct": -8.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.300571+00:00", "fingerprint": "e28e83ed99c34d4a8f469035" }, "review": null, "ledger": { "seq": 312, "hash": "52fb61c4d0bc4ea0d69229fd028d32eb986693cf3c6db9a05cd877f1e24255a9", "prev_hash": "f07fe4bec94952ce56134292bdd38ea5625abd88f8320a65e5cfad3030a2317c", "recorded_at": "2026-07-07T05:26:20+00:00" } }, { "txn_id": "BB-RNW-0170", "assessment_id": "ra_d604ff617e0c461f", "occurred_at": "2026-07-07T14:28:37+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_3310d38f3a" }, "output": { "decision": "rate", "score": 0.722, "reason_codes": [], "premium_delta_pct": 10.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-08T14:28:37+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-07, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-07-08 with specific reasons. RegBench applied 10 rule(s) in force on 2026-07-07 and recorded 2 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.943227+00:00", "fingerprint": "fa4a1bb5964b672c6077f9b3" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 106.5, "reviewer_role": "compliance", "resolved_at": "2026-07-12T00:57:27.713919+00:00" }, "ledger": { "seq": 170, "hash": "547da407234726de9aca5bf0504836237ad57be1984896e37632833391843323", "prev_hash": "9fc2b501ba4c44b73e9f15fb74a3624810d98aac201ed8e3c9084589eebf55b1", "recorded_at": "2026-07-07T14:30:07+00:00" } }, { "txn_id": "BB-RNW-0196", "assessment_id": "ra_46283cccf4b444c4", "occurred_at": "2026-07-07T20:55:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_35cc31b44d" }, "output": { "decision": "rate", "score": 0.251, "reason_codes": [], "premium_delta_pct": 10.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-12T20:55:04+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-12 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-07 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.000404+00:00", "fingerprint": "2b186cdb6ebb3244158bc0b1" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 196, "hash": "a015ad58031f3a3891cebf59d7d934f2a731e95768ec565c41552d68d327cade", "prev_hash": "eaea6173a4f94a44436a6b5c91e85d0952921dae9f48cc2f78c4e5c27f2a181b", "recorded_at": "2026-07-07T20:56:57+00:00" } }, { "txn_id": "BB-RNW-0313", "assessment_id": "ra_73d8c8843c1f46d0", "occurred_at": "2026-07-09T03:27:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a43f47c229" }, "output": { "decision": "rate", "score": 0.84, "reason_codes": [], "premium_delta_pct": -1.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.302288+00:00", "fingerprint": "9f4ed58f50255d8e6e66f56a" }, "review": null, "ledger": { "seq": 313, "hash": "256023ea2f2c2a6054c832e52e3a4c5551cec8f4a86502df7e2fc9051ed58482", "prev_hash": "52fb61c4d0bc4ea0d69229fd028d32eb986693cf3c6db9a05cd877f1e24255a9", "recorded_at": "2026-07-09T03:31:34+00:00" } }, { "txn_id": "BB-RNW-0152", "assessment_id": "ra_bd109fbf306f4427", "occurred_at": "2026-07-09T17:11:14+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e314eb3424" }, "output": { "decision": "rate", "score": 0.493, "reason_codes": [], "premium_delta_pct": 4.4 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.902162+00:00", "fingerprint": "b8b5c6467686f02164070501" }, "review": null, "ledger": { "seq": 152, "hash": "acf636f03334ea323d63d397bd9424a88718010ed43035246f9ca74fd8ac0490", "prev_hash": "46cc2e62b569c6d7f39f893ae9f0ad307d555eee57d02fba95e3deb0978729ac", "recorded_at": "2026-07-09T17:11:59+00:00" } }, { "txn_id": "BB-RNW-0202", "assessment_id": "ra_2327729b714d4719", "occurred_at": "2026-07-09T22:20:06+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7507287786" }, "output": { "decision": "rate", "score": 0.426, "reason_codes": [], "premium_delta_pct": 1.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.014932+00:00", "fingerprint": "72384bd274ccd135e0235e44" }, "review": null, "ledger": { "seq": 202, "hash": "ffac5dd426af20c81f7fdf555c13af88bc47b607cac8017701303cd4f92bd955", "prev_hash": "b5fa786ccbc1427aec251f3b0b2fea5b01d9d62d94e5f0a4d00278bcd7ba043f", "recorded_at": "2026-07-09T22:22:09+00:00" } }, { "txn_id": "BB-RNW-0296", "assessment_id": "ra_71cbb8b34c5a40ca", "occurred_at": "2026-07-10T06:56:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4016ea86af" }, "output": { "decision": "rate", "score": 0.512, "reason_codes": [], "premium_delta_pct": -2.4 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.261796+00:00", "fingerprint": "84574c5d5046274cd97c2dc6" }, "review": null, "ledger": { "seq": 296, "hash": "6101712e9a9d58c24db26af66b1ddc75bcd6a572ed6aaceb5cddadbe868746e0", "prev_hash": "a71e0a7dbcc4c7e6c5bb92fc99c7b70068ac74c6e976b33a11f1a89e644f75ed", "recorded_at": "2026-07-10T07:00:33+00:00" } }, { "txn_id": "BB-RNW-0243", "assessment_id": "ra_aa8e78c6867b45fc", "occurred_at": "2026-07-10T09:57:22+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b6fefb1aa6" }, "output": { "decision": "rate", "score": 0.448, "reason_codes": [], "premium_delta_pct": -11.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.116249+00:00", "fingerprint": "ce515b0c165d2bc98d48af7c" }, "review": null, "ledger": { "seq": 243, "hash": "41b558ed4966955e26c538ba3a20ae48e2a80467538cfb510dd43e8c4f6c46b2", "prev_hash": "97f22c1330b5fc0bf0bffd6e7deafd323eb82db0f9d1fa9734c57011b829da6c", "recorded_at": "2026-07-10T09:59:01+00:00" } }, { "txn_id": "BB-RNW-0273", "assessment_id": "ra_7b9e8b5d5dcd44f9", "occurred_at": "2026-07-11T01:12:34+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_15f1efbbf9" }, "output": { "decision": "rate", "score": 0.473, "reason_codes": [], "premium_delta_pct": 0.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-11, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-11 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.195238+00:00", "fingerprint": "c501794ab8989ab9585b8a9c" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 16.5, "reviewer_role": "compliance", "resolved_at": "2026-07-11T17:44:46.325531+00:00" }, "ledger": { "seq": 273, "hash": "6e6d9fe1f30eb6b54cab6c9288319476876a569dbf36e02f16a03832ad217bfc", "prev_hash": "79d98ac820a2b4ef30de3c6708cceded119f6807c3f5e080b15d48fc1aa262e5", "recorded_at": "2026-07-11T01:16:03+00:00" } }, { "txn_id": "BB-RNW-0208", "assessment_id": "ra_571208c37c614173", "occurred_at": "2026-07-12T11:27:17+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_95488faa57" }, "output": { "decision": "rate", "score": 0.253, "reason_codes": [], "premium_delta_pct": 1.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-12, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.030939+00:00", "fingerprint": "87425a7c792daa89fecf2bd0" }, "review": null, "ledger": { "seq": 208, "hash": "7ab739b8a72d9fe07c1d8cc7921439918f30cfd816d7ea0a8cafde3adcc53643", "prev_hash": "790f76f32c1e37c4e55a50ea2f3436230f65ae29b1a9ec9cde3aea008bba0f66", "recorded_at": "2026-07-12T11:29:36+00:00" } }, { "txn_id": "BB-RNW-0168", "assessment_id": "ra_9352c796b3dc4f46", "occurred_at": "2026-07-12T19:42:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0a74be6188" }, "output": { "decision": "rate", "score": 0.29, "reason_codes": [], "premium_delta_pct": 6.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-12, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.937380+00:00", "fingerprint": "bb882bb6ada56338058dba17" }, "review": null, "ledger": { "seq": 168, "hash": "8cfe1e1f5308589b0a35c7100cd25592e701846f521b8c64d4a7a2bb119d00c3", "prev_hash": "fb60aa0a7e6f0dfaeb0b627c4cd74f16d856dcbdbae481e52ebc8a10f8a8e780", "recorded_at": "2026-07-12T19:44:11+00:00" } }, { "txn_id": "BB-RNW-0227", "assessment_id": "ra_a9e18ea107094afb", "occurred_at": "2026-07-14T02:40:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_737d36e087" }, "output": { "decision": "rate", "score": 0.643, "reason_codes": [], "premium_delta_pct": 8.8 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.076076+00:00", "fingerprint": "dc3784135cae99c117f67294" }, "review": null, "ledger": { "seq": 227, "hash": "babc8f48561ef7e4c07d19cb371febe29a506ea422726a2ca3c74a9a3a1db462", "prev_hash": "fcac7275b64e9ea1b321e185fc552eea1dcc1b71f058ebf40b8b56dce2b72b96", "recorded_at": "2026-07-14T02:44:31+00:00" } }, { "txn_id": "BB-RNW-0245", "assessment_id": "ra_a8b2e333e0354a5a", "occurred_at": "2026-07-14T06:11:15+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d24fe72662" }, "output": { "decision": "rate", "score": 0.515, "reason_codes": [], "premium_delta_pct": 19.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-17T06:11:15+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-17 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.119941+00:00", "fingerprint": "db0b7a28bc607315c523402f" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 11.2, "reviewer_role": "compliance", "resolved_at": "2026-07-14T17:22:23.296292+00:00" }, "ledger": { "seq": 245, "hash": "ff4290daaf6d1115fa760e42bd44bec771900fb8dd61e3e21e4fc9d42a32c877", "prev_hash": "d879d362fb945e42e1c1e6f7aa24372ba7ecf913b18a386c2310c68b7c9a735b", "recorded_at": "2026-07-14T06:13:13+00:00" } }, { "txn_id": "BB-RNW-0206", "assessment_id": "ra_9205a35a59df4cd5", "occurred_at": "2026-07-14T09:56:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_ab0cb78a00" }, "output": { "decision": "rate", "score": 0.404, "reason_codes": [], "premium_delta_pct": 1.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.027251+00:00", "fingerprint": "2644116301c8230ce096388e" }, "review": null, "ledger": { "seq": 206, "hash": "8199733c16df8c6f0b6f787c0ba7167a27c5c159b7467c0e0b7587243aab0e83", "prev_hash": "8043b12a9f1c6718eb61642368462bdc1ada0e0215912d7edc514b75f9546ccb", "recorded_at": "2026-07-14T09:58:48+00:00" } }, { "txn_id": "BB-RNW-0241", "assessment_id": "ra_f96304d6fce041b2", "occurred_at": "2026-07-14T11:15:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d21400d496" }, "output": { "decision": "rate", "score": 0.833, "reason_codes": [], "premium_delta_pct": 7.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.112469+00:00", "fingerprint": "3d224b766ab8f6fa1dd7807a" }, "review": null, "ledger": { "seq": 241, "hash": "a087a56849041ec9c9e181c8dea1806d4e5f956cdc636183d5708393583cf56a", "prev_hash": "3e8334236a9df5ea5295cc8f600a8af9ecd5ef403a631bb371275302300f2db3", "recorded_at": "2026-07-14T11:17:31+00:00" } }, { "txn_id": "BB-RNW-0201", "assessment_id": "ra_e2c8dbfe27544b16", "occurred_at": "2026-07-14T17:34:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4a13e09a85" }, "output": { "decision": "rate", "score": 0.223, "reason_codes": [], "premium_delta_pct": 11.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-17T17:34:51+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-17 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.012996+00:00", "fingerprint": "4e2d0249fbc866143710553b" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 74.9, "reviewer_role": "compliance", "resolved_at": "2026-07-17T20:34:14.422467+00:00" }, "ledger": { "seq": 201, "hash": "b5fa786ccbc1427aec251f3b0b2fea5b01d9d62d94e5f0a4d00278bcd7ba043f", "prev_hash": "5235b613ca27263508f9d34dbdca0f01942a399a20d4dbd5ab75e4a53bfa5a21", "recorded_at": "2026-07-14T17:37:58+00:00" } }, { "txn_id": "BB-RNW-0204", "assessment_id": "ra_a1a86853c43d459d", "occurred_at": "2026-07-14T21:58:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_549bee0f5e" }, "output": { "decision": "rate", "score": 0.745, "reason_codes": [], "premium_delta_pct": 10.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-15T21:58:32+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-15 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.019515+00:00", "fingerprint": "90eceb788bb5214d50f9c13c" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 28.3, "reviewer_role": "compliance", "resolved_at": "2026-07-16T02:18:28.323713+00:00" }, "ledger": { "seq": 204, "hash": "89f67ef193f5b066e1368de2a19fcbe36c609dc12026f9c5d289232bc5697504", "prev_hash": "9a8f1bcc760cddf798d8b5f5bf4651566a6c5f1d4ae78996d8582d883e765012", "recorded_at": "2026-07-14T21:59:25+00:00" } }, { "txn_id": "BB-RNW-0234", "assessment_id": "ra_02ad99b17c424552", "occurred_at": "2026-07-15T16:32:40+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f84e8e3b56" }, "output": { "decision": "rate", "score": 0.297, "reason_codes": [], "premium_delta_pct": 6.7 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-15, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.094769+00:00", "fingerprint": "a0336ee93e77a453c4fdab92" }, "review": null, "ledger": { "seq": 234, "hash": "5d1791bf99bbc265d012211508b23417e6283b913b933fe741e538bfd0ed73ab", "prev_hash": "1e7d10fea127bad4abaf023cd419ea20155ecbbc1557c49ff67d751d91ac693d", "recorded_at": "2026-07-15T16:36:39+00:00" } }, { "txn_id": "BB-RNW-0292", "assessment_id": "ra_bb9b4db03cd84b44", "occurred_at": "2026-07-16T19:57:14+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e4205d70b6" }, "output": { "decision": "rate", "score": 0.747, "reason_codes": [], "premium_delta_pct": 13.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-21T19:57:14+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-16, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-21 with specific reasons. RegBench applied 10 rule(s) in force on 2026-07-16 and recorded 3 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.250696+00:00", "fingerprint": "ec51365e571c606cc1d02faf" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 32.5, "reviewer_role": "compliance", "resolved_at": "2026-07-18T04:32:52.224144+00:00" }, "ledger": { "seq": 292, "hash": "4f8c51a82c14d866638806b671594d196e19a9e33b085aea6a6db6478d3cb617", "prev_hash": "46f5330dedfdeb3d2f55309551a566a8008c1c7fae09e480a01611c99782a7e5", "recorded_at": "2026-07-16T20:00:06+00:00" } }, { "txn_id": "BB-RNW-0194", "assessment_id": "ra_be7074527b41485e", "occurred_at": "2026-07-17T10:38:35+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_174959b514" }, "output": { "decision": "rate", "score": 0.351, "reason_codes": [], "premium_delta_pct": 13.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-17, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-07-17 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.996246+00:00", "fingerprint": "322cd1c38bb4296873ed0f17" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 4.7, "reviewer_role": "compliance", "resolved_at": "2026-07-17T15:21:43.855042+00:00" }, "ledger": { "seq": 194, "hash": "a580c3c55cd0aa94aa6aaed1031a1178addf50938ac4b74a8f1b306a006bfec0", "prev_hash": "e3028825e1aa49d37285b4f48884bd2e20dc330bee7ee5c72eb938ab8f218461", "recorded_at": "2026-07-17T10:41:37+00:00" } }, { "txn_id": "BB-RNW-0251", "assessment_id": "ra_d02fafc68b6d4a08", "occurred_at": "2026-07-17T20:16:50+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0b76c5660a" }, "output": { "decision": "rate", "score": 0.643, "reason_codes": [], "premium_delta_pct": 7.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-17, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.140608+00:00", "fingerprint": "83a056cae6f3778576d491ad" }, "review": null, "ledger": { "seq": 251, "hash": "b11dac1da5aa740f4b3ea4f91515a423f8378cd07bb5cb13070a813a5b57f79f", "prev_hash": "f3542fb901ccfa5e79075bf067e77f4398439a8f3c35242cd0e094cd3616447e", "recorded_at": "2026-07-17T20:20:04+00:00" } }, { "txn_id": "BB-RNW-0181", "assessment_id": "ra_240bdde8a92f46bd", "occurred_at": "2026-07-18T03:02:53+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9c39525c17" }, "output": { "decision": "rate", "score": 0.438, "reason_codes": [], "premium_delta_pct": 1.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-18, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.967680+00:00", "fingerprint": "699d2bcfd0d864e8c6d668fd" }, "review": null, "ledger": { "seq": 181, "hash": "8cb5401e9949b82f0df66e4bf764f79bbf341b33542dede477cf4e858a2b2cbd", "prev_hash": "a82086851eb714053de9d49db92d090727e2c514f9b36921053e3e6ce365a259", "recorded_at": "2026-07-18T03:05:19+00:00" } }, { "txn_id": "BB-RNW-0198", "assessment_id": "ra_2033891307b9456a", "occurred_at": "2026-07-18T15:10:02+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_aa5f152e63" }, "output": { "decision": "rate", "score": 0.874, "reason_codes": [], "premium_delta_pct": 0.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-18, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-18 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.007264+00:00", "fingerprint": "bff10d66b814f2ad55ca00f6" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 7.2, "reviewer_role": "compliance", "resolved_at": "2026-07-18T22:24:53.595994+00:00" }, "ledger": { "seq": 198, "hash": "e1b7358f471ccd79506c012700c25abf5a72984ad776f26b42ff227a9ad24b43", "prev_hash": "d86243ff7da848e45ac7ddb8dba8beec48eb5384447ea15c9aedfbec23923dbb", "recorded_at": "2026-07-18T15:13:26+00:00" } }, { "txn_id": "BB-RNW-0244", "assessment_id": "ra_6270e1fd877b4f51", "occurred_at": "2026-07-19T02:54:21+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d1f8e04c40" }, "output": { "decision": "rate", "score": 0.212, "reason_codes": [], "premium_delta_pct": 0.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.117998+00:00", "fingerprint": "aa198d77549144719843a811" }, "review": null, "ledger": { "seq": 244, "hash": "d879d362fb945e42e1c1e6f7aa24372ba7ecf913b18a386c2310c68b7c9a735b", "prev_hash": "41b558ed4966955e26c538ba3a20ae48e2a80467538cfb510dd43e8c4f6c46b2", "recorded_at": "2026-07-19T02:56:59+00:00" } }, { "txn_id": "BB-RNW-0261", "assessment_id": "ra_fb7d65f51bec4f4b", "occurred_at": "2026-07-19T13:49:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f3394a47e0" }, "output": { "decision": "rate", "score": 0.887, "reason_codes": [], "premium_delta_pct": 2.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.162425+00:00", "fingerprint": "0042b64700d0e0f31187391c" }, "review": null, "ledger": { "seq": 261, "hash": "c1dbe2646796c411319e148fb38a3b09878bc7157e2813b83f727df5436fde9b", "prev_hash": "70b4451d23c2e748c38007be902f56c25d5203deae9a7dfb94f1f415923bb9da", "recorded_at": "2026-07-19T13:53:24+00:00" } }, { "txn_id": "BB-RNW-0268", "assessment_id": "ra_93434ea49c8e470b", "occurred_at": "2026-07-20T03:00:58+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8979f50744" }, "output": { "decision": "rate", "score": 0.237, "reason_codes": [], "premium_delta_pct": 0.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-20, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.181733+00:00", "fingerprint": "9c1ffef1f717392fa3bad28b" }, "review": null, "ledger": { "seq": 268, "hash": "3a5a837d960a6204661d74b1f438127303fb6651172b68347642fde20cedc84b", "prev_hash": "50d648af26f6bf26b363c29a7b3765377d55f194a39ec59eb5d9e3777c4f4dec", "recorded_at": "2026-07-20T03:01:45+00:00" } }, { "txn_id": "BB-RNW-0291", "assessment_id": "ra_24463966a0fe4c53", "occurred_at": "2026-07-20T09:43:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_220b4fd025" }, "output": { "decision": "rate", "score": 0.707, "reason_codes": [], "premium_delta_pct": 12.3 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-23T09:43:32+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-20, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-07-23 with reasons stated. RegBench applied 11 rule(s) in force on 2026-07-20 and recorded 2 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.249029+00:00", "fingerprint": "b4b8873de5da0ed66807d31f" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 11.8, "reviewer_role": "compliance", "resolved_at": "2026-07-20T21:35:19.475353+00:00" }, "ledger": { "seq": 291, "hash": "46f5330dedfdeb3d2f55309551a566a8008c1c7fae09e480a01611c99782a7e5", "prev_hash": "5faf91d2d7a1249ed77f9e20c871ae155c5400d795bb8a160f8f1a5eae0f9f65", "recorded_at": "2026-07-20T09:47:05+00:00" } }, { "txn_id": "BB-RNW-0253", "assessment_id": "ra_aa9b2c0861b34d65", "occurred_at": "2026-07-23T01:12:22+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_aefa20e65e" }, "output": { "decision": "rate", "score": 0.803, "reason_codes": [], "premium_delta_pct": -10.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.145603+00:00", "fingerprint": "32fd35724785ae9b729a5c1f" }, "review": null, "ledger": { "seq": 253, "hash": "bc248fac7c707679aed58ae5b8d5853becf0e6e0d919a6d304e6a3a167776ff2", "prev_hash": "bb8f0a88ccf294f0a2495f1d3a5e039e2665f29aa671f1f8250a5dee766771ef", "recorded_at": "2026-07-23T01:13:35+00:00" } }, { "txn_id": "BB-RNW-0175", "assessment_id": "ra_7fcf106295354cfe", "occurred_at": "2026-07-23T12:34:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_04914122a7" }, "output": { "decision": "rate", "score": 0.556, "reason_codes": [], "premium_delta_pct": 8.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.951574+00:00", "fingerprint": "739f2da1caf641a006279742" }, "review": null, "ledger": { "seq": 175, "hash": "26e0424cc0b689e76d53f88a870a012af2e00617e6a59cb42d58aeb7918db904", "prev_hash": "070c804d27e15500d6a4a68e665a6dc8a57b209f9324e774d3eee4555a681809", "recorded_at": "2026-07-23T12:37:30+00:00" } }, { "txn_id": "BB-RNW-0305", "assessment_id": "ra_bcda523863de4605", "occurred_at": "2026-07-23T22:36:44+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a7e06e51e8" }, "output": { "decision": "rate", "score": 0.557, "reason_codes": [], "premium_delta_pct": 7.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.283955+00:00", "fingerprint": "739f2da1caf641a006279742" }, "review": null, "ledger": { "seq": 305, "hash": "1a09e27c593e14f9791da5c05564a90c01cc5b5e6acabf2d9ff370551471eb0b", "prev_hash": "c139c7d282672f418fb5999aa3faba2d6c7909d8e613f6f9b657deed50cf52f8", "recorded_at": "2026-07-23T22:37:34+00:00" } }, { "txn_id": "BB-RNW-0257", "assessment_id": "ra_29040c8c830849eb", "occurred_at": "2026-07-24T04:36:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4a0bbc8157" }, "output": { "decision": "rate", "score": 0.68, "reason_codes": [], "premium_delta_pct": 20.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-25T04:36:53+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-07-25 with reasons stated. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded 2 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.153035+00:00", "fingerprint": "9216296f5e0352045d8e7802" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 1.9, "reviewer_role": "compliance", "resolved_at": "2026-07-24T06:35:02.863243+00:00" }, "ledger": { "seq": 257, "hash": "ed26cae9e4f8d7276ae83f316c12e6c90eec138bb907e50ea001c3dd071405e6", "prev_hash": "c20aaf83ebc64ea615faa1a8a781e160bed3fdb3d59bb58ea08296ca9b973838", "recorded_at": "2026-07-24T04:40:26+00:00" } }, { "txn_id": "BB-RNW-0156", "assessment_id": "ra_5ee788bddeb54a28", "occurred_at": "2026-07-24T06:59:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d022a5efaa" }, "output": { "decision": "rate", "score": 0.619, "reason_codes": [], "premium_delta_pct": 5.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.910665+00:00", "fingerprint": "d1b12e7d8e5fe13cf6513c22" }, "review": null, "ledger": { "seq": 156, "hash": "2dcc88a2a3d83e16bad82cf574bfc894aadccfdf6a677b4aadf4105aa1448cc4", "prev_hash": "f1cb04da97a9acca33bab9e9b9d72a537307fb5327053e1417dd48d357f9c67c", "recorded_at": "2026-07-24T07:00:50+00:00" } }, { "txn_id": "BB-RNW-0164", "assessment_id": "ra_f6d3189bef0b4081", "occurred_at": "2026-07-24T17:13:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_1e50145f2e" }, "output": { "decision": "rate", "score": 0.575, "reason_codes": [], "premium_delta_pct": -3.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.930396+00:00", "fingerprint": "d1b12e7d8e5fe13cf6513c22" }, "review": null, "ledger": { "seq": 164, "hash": "aa18133999920e16ee7762c5c5d3f0015c50b50d25d90b330807840fd85148cf", "prev_hash": "899968a82b0bd419dea801f8ec2cceaa084df2a43aa0c9db7aff6e00fdf42220", "recorded_at": "2026-07-24T17:17:03+00:00" } }, { "txn_id": "BB-RNW-0311", "assessment_id": "ra_9ae81b84fd4b4597", "occurred_at": "2026-07-25T18:02:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f108b1d1c1" }, "output": { "decision": "rate", "score": 0.399, "reason_codes": [], "premium_delta_pct": 12.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-30T18:02:11+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-30 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-25 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.298897+00:00", "fingerprint": "0d276bf7e6edb90bde5d335f" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 29.6, "reviewer_role": "compliance", "resolved_at": "2026-07-26T23:37:32.304060+00:00" }, "ledger": { "seq": 311, "hash": "f07fe4bec94952ce56134292bdd38ea5625abd88f8320a65e5cfad3030a2317c", "prev_hash": "8bf70166591b81e1fd31772be3645d300fbdebd78427104a11c849fc0e04c09d", "recorded_at": "2026-07-25T18:02:41+00:00" } }, { "txn_id": "BB-RNW-0167", "assessment_id": "ra_a62f9396f6f24398", "occurred_at": "2026-07-26T22:19:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b0187324f1" }, "output": { "decision": "rate", "score": 0.266, "reason_codes": [], "premium_delta_pct": 0.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-26, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.936015+00:00", "fingerprint": "1262b2b7fc3ce3ee7f720636" }, "review": null, "ledger": { "seq": 167, "hash": "fb60aa0a7e6f0dfaeb0b627c4cd74f16d856dcbdbae481e52ebc8a10f8a8e780", "prev_hash": "691e97863e85d17572159592d00bf72aaace3fba4a2b5c6ad876126d2a51b4b5", "recorded_at": "2026-07-26T22:19:35+00:00" } }, { "txn_id": "BB-RNW-0174", "assessment_id": "ra_e894ecc59fc24e20", "occurred_at": "2026-07-27T15:23:12+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a699e61a18" }, "output": { "decision": "rate", "score": 0.282, "reason_codes": [], "premium_delta_pct": 3.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-27, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.950133+00:00", "fingerprint": "983daae17a44f794300131c5" }, "review": null, "ledger": { "seq": 174, "hash": "070c804d27e15500d6a4a68e665a6dc8a57b209f9324e774d3eee4555a681809", "prev_hash": "d87ef0f80fdae3fc8917a9381b96df3d416d6111febdd635110e13370f867a82", "recorded_at": "2026-07-27T15:24:39+00:00" } }, { "txn_id": "BB-RNW-0165", "assessment_id": "ra_887c1dc5d7dd41f2", "occurred_at": "2026-07-28T19:23:16+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_dd57c9906a" }, "output": { "decision": "rate", "score": 0.506, "reason_codes": [], "premium_delta_pct": -11.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-28, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.932505+00:00", "fingerprint": "440ccf1723d090994058c728" }, "review": null, "ledger": { "seq": 165, "hash": "937655b5ed3676a40bc371a8f4fa030c3b96d9f0de641f0dd1b768540cdf4062", "prev_hash": "aa18133999920e16ee7762c5c5d3f0015c50b50d25d90b330807840fd85148cf", "recorded_at": "2026-07-28T19:27:12+00:00" } }, { "txn_id": "BB-RNW-0280", "assessment_id": "ra_41c8903dc1e84e6a", "occurred_at": "2026-07-29T22:03:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_2a59961ccb" }, "output": { "decision": "rate", "score": 0.497, "reason_codes": [], "premium_delta_pct": 15.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-03T22:03:11+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-29, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-03 with reasons stated. RegBench applied 11 rule(s) in force on 2026-07-29 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.218917+00:00", "fingerprint": "371465df984caa467076b3c1" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 4.4, "reviewer_role": "compliance", "resolved_at": "2026-07-30T02:28:27.397377+00:00" }, "ledger": { "seq": 280, "hash": "46bbc154043ac8a51d8163cec5c297125590623295044c26b5544fb9d11c0ba6", "prev_hash": "062542f6c06d6ed78bcf178cd0fda9def5b82065d23b16020e3f0e35f71332ec", "recorded_at": "2026-07-29T22:04:17+00:00" } }, { "txn_id": "BB-RNW-0199", "assessment_id": "ra_2c4d08ae29de4dac", "occurred_at": "2026-07-30T00:24:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_2e9da59210" }, "output": { "decision": "rate", "score": 0.87, "reason_codes": [], "premium_delta_pct": 1.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.009456+00:00", "fingerprint": "8ec884a9d892124a09dfa736" }, "review": null, "ledger": { "seq": 199, "hash": "27bf182dc9ed7236c2be3bab639a4fac7b884d687223b04f9a1f561ca0116716", "prev_hash": "e1b7358f471ccd79506c012700c25abf5a72984ad776f26b42ff227a9ad24b43", "recorded_at": "2026-07-30T00:26:22+00:00" } }, { "txn_id": "BB-RNW-0153", "assessment_id": "ra_dc8d989442d9458d", "occurred_at": "2026-07-30T01:54:48+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9ac6ae807d" }, "output": { "decision": "rate", "score": 0.437, "reason_codes": [], "premium_delta_pct": 2.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-30, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-30 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.903595+00:00", "fingerprint": "b2b8f21fccc7a7320d9aad95" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 6.6, "reviewer_role": "compliance", "resolved_at": "2026-07-30T08:35:57.298654+00:00" }, "ledger": { "seq": 153, "hash": "0a6907ce64f5ad8c9da5f6671f8f232689c4794fc1c0a1f3a05f948305773a2c", "prev_hash": "acf636f03334ea323d63d397bd9424a88718010ed43035246f9ca74fd8ac0490", "recorded_at": "2026-07-30T01:58:24+00:00" } }, { "txn_id": "BB-RNW-0192", "assessment_id": "ra_8d7191c10c0a4b6b", "occurred_at": "2026-07-30T18:16:57+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_da46ca3a78" }, "output": { "decision": "rate", "score": 0.244, "reason_codes": [], "premium_delta_pct": -0.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-30, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.992145+00:00", "fingerprint": "c1b4c8c40972d0d9d1c5c64d" }, "review": null, "ledger": { "seq": 192, "hash": "5677ae0a5560fcfc89709eb7397c6dd0e03bb99ba43053350904afd9b0c450bb", "prev_hash": "6350ae4e8d5eca4a0b5d66f2dee8be38afdd53fce46a2ad6ba2361b961250f57", "recorded_at": "2026-07-30T18:19:55+00:00" } }, { "txn_id": "BB-RNW-0277", "assessment_id": "ra_57ba9818459e4d4f", "occurred_at": "2026-08-01T05:19:18+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4d22dab971" }, "output": { "decision": "rate", "score": 0.378, "reason_codes": [], "premium_delta_pct": 3.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-01, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-01 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.207817+00:00", "fingerprint": "8555db097227a337666a4d32" }, "review": null, "ledger": { "seq": 277, "hash": "ffac8e1a2e4e388aca9357ed544423d0a8d9453d4a3c3219aee994137f94ccc1", "prev_hash": "0ed499a75c9a79b7342a4c776a4bbe9e0b4eeea1cfc23abeb98f98a33c0eed55", "recorded_at": "2026-08-01T05:20:17+00:00" } }, { "txn_id": "BB-RNW-0160", "assessment_id": "ra_f9c3e3ca0f79419f", "occurred_at": "2026-08-01T06:30:17+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4360457424" }, "output": { "decision": "rate", "score": 0.891, "reason_codes": [], "premium_delta_pct": -0.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-01, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-01 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.920577+00:00", "fingerprint": "baa532b6402373a0028ce3bb" }, "review": null, "ledger": { "seq": 160, "hash": "e3b0fc1c3b9ccfb1f00b1392eacd7159248a07c4cf8a6434a37a89b78fae39c1", "prev_hash": "3827e70cb8010bf5cadb068138e7515c98628586bb24c6b63c7ff495d7c2554e", "recorded_at": "2026-08-01T06:31:01+00:00" } }, { "txn_id": "BB-RNW-0197", "assessment_id": "ra_6e05f7d71efa4c94", "occurred_at": "2026-08-02T07:21:58+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d5772b03fb" }, "output": { "decision": "rate", "score": 0.283, "reason_codes": [], "premium_delta_pct": -18.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-02, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.003277+00:00", "fingerprint": "54c4bd22d66ea5747defc5b3" }, "review": null, "ledger": { "seq": 197, "hash": "d86243ff7da848e45ac7ddb8dba8beec48eb5384447ea15c9aedfbec23923dbb", "prev_hash": "a015ad58031f3a3891cebf59d7d934f2a731e95768ec565c41552d68d327cade", "recorded_at": "2026-08-02T07:25:05+00:00" } }, { "txn_id": "BB-RNW-0154", "assessment_id": "ra_317409ebec1b4ef8", "occurred_at": "2026-08-02T22:10:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_98342aff74" }, "output": { "decision": "rate", "score": 0.247, "reason_codes": [], "premium_delta_pct": 4.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.907793+00:00", "fingerprint": "cae8598a853960c31d916cc7" }, "review": null, "ledger": { "seq": 154, "hash": "281b3781cf051db7b74b67007cd07fa66f58b47ea4d0d5c9d396bcd138a5c8f9", "prev_hash": "0a6907ce64f5ad8c9da5f6671f8f232689c4794fc1c0a1f3a05f948305773a2c", "recorded_at": "2026-08-02T22:13:54+00:00" } }, { "txn_id": "BB-RNW-0161", "assessment_id": "ra_1ac33ba234a64e7c", "occurred_at": "2026-08-03T00:00:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_06929323c3" }, "output": { "decision": "rate", "score": 0.322, "reason_codes": [], "premium_delta_pct": 14.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-05T00:00:45+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-05 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-03 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.925194+00:00", "fingerprint": "1c7388d7826c05619f68daf6" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 5.3, "reviewer_role": "compliance", "resolved_at": "2026-08-03T05:22:43.393218+00:00" }, "ledger": { "seq": 161, "hash": "8c3db1a3515723b910fabce35dc212065ea8ef52a8c3b667218b0a9cab68bbbb", "prev_hash": "e3b0fc1c3b9ccfb1f00b1392eacd7159248a07c4cf8a6434a37a89b78fae39c1", "recorded_at": "2026-08-03T00:01:49+00:00" } }, { "txn_id": "BB-RNW-0246", "assessment_id": "ra_10c1034ea20a4452", "occurred_at": "2026-08-03T00:24:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_6626332853" }, "output": { "decision": "rate", "score": 0.763, "reason_codes": [], "premium_delta_pct": -13.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.126452+00:00", "fingerprint": "d02a53b7e8f952b48ee9f88b" }, "review": null, "ledger": { "seq": 246, "hash": "af91042b36ccf89716243850069146daede522b91a9daff393e22ead57f459da", "prev_hash": "ff4290daaf6d1115fa760e42bd44bec771900fb8dd61e3e21e4fc9d42a32c877", "recorded_at": "2026-08-03T00:26:28+00:00" } }, { "txn_id": "BB-RNW-0188", "assessment_id": "ra_212a2a61bec54a21", "occurred_at": "2026-08-05T02:05:37+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9408ad0fc3" }, "output": { "decision": "rate", "score": 0.487, "reason_codes": [], "premium_delta_pct": 3.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-05, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.981219+00:00", "fingerprint": "f746fecf368434c9c6df5577" }, "review": null, "ledger": { "seq": 188, "hash": "3a2d67d4f66284b30eaaaa853b30b550b2ecb0aa32727040a5222c8d4c7445ad", "prev_hash": "f54f446037ed3847a40ff072c2568fc0def57b422aeb980aa4dc455271bdd05b", "recorded_at": "2026-08-05T02:07:14+00:00" } }, { "txn_id": "BB-RNW-0271", "assessment_id": "ra_bda3e7c0b3d843d2", "occurred_at": "2026-08-06T07:59:37+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_c3e2d0041c" }, "output": { "decision": "rate", "score": 0.64, "reason_codes": [], "premium_delta_pct": 0.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.190940+00:00", "fingerprint": "75da2da987153822500bae40" }, "review": null, "ledger": { "seq": 271, "hash": "82f57953fa6cddece4221eedc3594579d46a9a5fcd6372cf0f25fb0522fd6698", "prev_hash": "3a9d1ea06784493184c143bece6de9f44e16b5f418ec02050a3fb6b3c5b9f769", "recorded_at": "2026-08-06T08:02:07+00:00" } }, { "txn_id": "BB-RNW-0266", "assessment_id": "ra_cc01dd90673e41a0", "occurred_at": "2026-08-06T10:27:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a237c6ead4" }, "output": { "decision": "rate", "score": 0.398, "reason_codes": [], "premium_delta_pct": 6.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.177521+00:00", "fingerprint": "3529b0975669da6007b8b14f" }, "review": null, "ledger": { "seq": 266, "hash": "4bdacd8d1f76cc0245d7f3e63544833feeea6793f7ac5598239057f08417a68b", "prev_hash": "14304034db2f312bc140c3694a96d641d04d36405b8583cf2f30bc654ba2df5f", "recorded_at": "2026-08-06T10:31:14+00:00" } }, { "txn_id": "BB-RNW-0287", "assessment_id": "ra_d53b210c4c48400a", "occurred_at": "2026-08-07T08:38:52+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9f997c3483" }, "output": { "decision": "rate", "score": 0.27, "reason_codes": [], "premium_delta_pct": -11.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-07, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.242156+00:00", "fingerprint": "d6e4d277e82fe7c4ed34a9bc" }, "review": null, "ledger": { "seq": 287, "hash": "1df44da920e402366268f37e68fef1832c7f4c920f9a31f71e92ce59746415e6", "prev_hash": "2bbd305a21832e7093ab51f62fd6476cbc49bb28623e498e8caaa5b091a9a1c5", "recorded_at": "2026-08-07T08:39:53+00:00" } }, { "txn_id": "BB-RNW-0237", "assessment_id": "ra_54f09754428146c3", "occurred_at": "2026-08-09T21:06:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_1fd47b66c2" }, "output": { "decision": "rate", "score": 0.604, "reason_codes": [], "premium_delta_pct": 6.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-09, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.101370+00:00", "fingerprint": "5afaa537aec8d60de7871f5d" }, "review": null, "ledger": { "seq": 237, "hash": "84b1a3d298d69315fa712b8533bea6f92922412747bc30f329068047dbd2b3b8", "prev_hash": "840ed64e6c8aba3caf714c42de61150610bd70ce8f401f32880b010280e64ba0", "recorded_at": "2026-08-09T21:09:40+00:00" } }, { "txn_id": "BB-RNW-0238", "assessment_id": "ra_448ca09f34474950", "occurred_at": "2026-08-10T03:32:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_82eb436d88" }, "output": { "decision": "rate", "score": 0.403, "reason_codes": [], "premium_delta_pct": -1.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-10, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.103168+00:00", "fingerprint": "dc680fde1ce6a69899897c26" }, "review": null, "ledger": { "seq": 238, "hash": "24159b593b9597fba47ea1db4217fe04d7cf27450d6e6a951f3eb38628f82118", "prev_hash": "84b1a3d298d69315fa712b8533bea6f92922412747bc30f329068047dbd2b3b8", "recorded_at": "2026-08-10T03:34:26+00:00" } }, { "txn_id": "BB-RNW-0215", "assessment_id": "ra_e59978ad628f4eca", "occurred_at": "2026-08-10T14:17:17+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_133f03a4ad" }, "output": { "decision": "rate", "score": 0.391, "reason_codes": [], "premium_delta_pct": 7.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-10, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.048108+00:00", "fingerprint": "dc680fde1ce6a69899897c26" }, "review": null, "ledger": { "seq": 215, "hash": "deaca80be5ebc4e9f98b0aab5b8b248061e0cc09e0bbbd480611e9b23077e353", "prev_hash": "5f1b640e27ece99b83e1d494309310d48ac03e8ecd7a24f849f47a9e6305775a", "recorded_at": "2026-08-10T14:18:31+00:00" } }, { "txn_id": "BB-RNW-0209", "assessment_id": "ra_49478d02194441c9", "occurred_at": "2026-08-11T20:16:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_61a4828e07" }, "output": { "decision": "rate", "score": 0.267, "reason_codes": [], "premium_delta_pct": 4.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-11, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.032633+00:00", "fingerprint": "a86912723a92c4640f55bd8b" }, "review": null, "ledger": { "seq": 209, "hash": "f890102a9f92d54e9f65c90535c8cdde5a613c8c5b6b7a9badacd6a20d6b6e87", "prev_hash": "7ab739b8a72d9fe07c1d8cc7921439918f30cfd816d7ea0a8cafde3adcc53643", "recorded_at": "2026-08-11T20:19:47+00:00" } }, { "txn_id": "BB-RNW-0177", "assessment_id": "ra_05ce9794c7dc4e3c", "occurred_at": "2026-08-11T23:05:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_cfb3ace279" }, "output": { "decision": "rate", "score": 0.343, "reason_codes": [], "premium_delta_pct": 14.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-16T23:05:25+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-11, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-16 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-11 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.958164+00:00", "fingerprint": "ef7ef05401eb747c07b6d1a7" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 15.7, "reviewer_role": "compliance", "resolved_at": "2026-08-12T14:49:00.119336+00:00" }, "ledger": { "seq": 177, "hash": "9060247a6c6f86225f0bc439289749e13c684747ee84b6d7ff0599134fa5c9e8", "prev_hash": "484915bf1dc6e1380e8c6fab2a34bd958c1b8abe73a46e9f00fbf22d45bfa0ec", "recorded_at": "2026-08-11T23:08:12+00:00" } }, { "txn_id": "BB-RNW-0260", "assessment_id": "ra_459bb1443c7a4d55", "occurred_at": "2026-08-12T13:46:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_28d19aeed5" }, "output": { "decision": "rate", "score": 0.88, "reason_codes": [], "premium_delta_pct": 2.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-12, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.160874+00:00", "fingerprint": "1252f8ca2fd2a18e3a6fdf4b" }, "review": null, "ledger": { "seq": 260, "hash": "70b4451d23c2e748c38007be902f56c25d5203deae9a7dfb94f1f415923bb9da", "prev_hash": "e409131134b224be2b1e0d2b0c77716dea0759202709e6918831ec16ce956118", "recorded_at": "2026-08-12T13:49:10+00:00" } }, { "txn_id": "BB-RNW-0247", "assessment_id": "ra_49b562ce50894d6e", "occurred_at": "2026-08-12T16:42:29+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_803c621e1d" }, "output": { "decision": "rate", "score": 0.438, "reason_codes": [], "premium_delta_pct": 8.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-12, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.129203+00:00", "fingerprint": "8ae9647902736162c1c10026" }, "review": null, "ledger": { "seq": 247, "hash": "4250e4565a9ad76cc990b9947c6b919d000acac199eeba9f5051e30c9e62bdb4", "prev_hash": "af91042b36ccf89716243850069146daede522b91a9daff393e22ead57f459da", "recorded_at": "2026-08-12T16:43:48+00:00" } }, { "txn_id": "BB-RNW-0176", "assessment_id": "ra_9dbee99d29e0402c", "occurred_at": "2026-08-14T02:21:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_cc5ffc9515" }, "output": { "decision": "rate", "score": 0.28, "reason_codes": [], "premium_delta_pct": 8.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.952962+00:00", "fingerprint": "a5d1b04b496cf935979bc8e4" }, "review": null, "ledger": { "seq": 176, "hash": "484915bf1dc6e1380e8c6fab2a34bd958c1b8abe73a46e9f00fbf22d45bfa0ec", "prev_hash": "26e0424cc0b689e76d53f88a870a012af2e00617e6a59cb42d58aeb7918db904", "recorded_at": "2026-08-14T02:25:29+00:00" } }, { "txn_id": "BB-RNW-0315", "assessment_id": "ra_c8acf5c88b6046d1", "occurred_at": "2026-08-14T20:47:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_54621d0823" }, "output": { "decision": "rate", "score": 0.243, "reason_codes": [], "premium_delta_pct": 18.8 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-19T20:47:57+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-08-19 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-14 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.308737+00:00", "fingerprint": "9a85b99900ab35b443597b77" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 18.5, "reviewer_role": "compliance", "resolved_at": "2026-08-15T15:21:39.589566+00:00" }, "ledger": { "seq": 315, "hash": "887d27ee022645df00ea56b974b5abb32107bb26b7ad37039680cf50c9ca6cd3", "prev_hash": "2696324d8b3a556eb6eecfa74c6b969900a67255f7d48b463ef3e4591e1d7a54", "recorded_at": "2026-08-14T20:51:47+00:00" } }, { "txn_id": "BB-RNW-0255", "assessment_id": "ra_c11c3d25d26243e2", "occurred_at": "2026-08-15T22:24:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_09744263f6" }, "output": { "decision": "rate", "score": 0.234, "reason_codes": [], "premium_delta_pct": 13.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-19T22:24:50+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-15, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-19 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-15 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.148859+00:00", "fingerprint": "72b413f56d9e9b3a169a6479" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 9.0, "reviewer_role": "compliance", "resolved_at": "2026-08-16T07:26:24.631439+00:00" }, "ledger": { "seq": 255, "hash": "d980903aecc9fd700af508768d5429c09dcda5a532b4cacfa94937f989f31fcb", "prev_hash": "fb51237e0434437f02b124eed32546740bf64f811d3d23b798e8a41514a50df8", "recorded_at": "2026-08-15T22:28:01+00:00" } }, { "txn_id": "BB-RNW-0303", "assessment_id": "ra_229b64d937864bf7", "occurred_at": "2026-08-17T08:16:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b7e53b8cd4" }, "output": { "decision": "rate", "score": 0.73, "reason_codes": [], "premium_delta_pct": -10.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-17, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-17 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.276470+00:00", "fingerprint": "73428188c0aff7280c10b078" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 38.7, "reviewer_role": "compliance", "resolved_at": "2026-08-18T22:56:51.648521+00:00" }, "ledger": { "seq": 303, "hash": "bd3d3558e57ec617747839ec4cf22a22b3795a5fec0d784e5068c83ee4a6535c", "prev_hash": "94e868cd051acba0a1f492e349ba8e872b781cc10e5ce26464c9dcf03a48d780", "recorded_at": "2026-08-17T08:16:44+00:00" } }, { "txn_id": "BB-RNW-0252", "assessment_id": "ra_7feab13edfb44ad9", "occurred_at": "2026-08-17T19:51:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_5d729a662a" }, "output": { "decision": "rate", "score": 0.839, "reason_codes": [], "premium_delta_pct": 12.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-22T19:51:56+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-17, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-22 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-17 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.143690+00:00", "fingerprint": "f32addb8991b9916e23172e0" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 252, "hash": "bb8f0a88ccf294f0a2495f1d3a5e039e2665f29aa671f1f8250a5dee766771ef", "prev_hash": "b11dac1da5aa740f4b3ea4f91515a423f8378cd07bb5cb13070a813a5b57f79f", "recorded_at": "2026-08-17T19:55:55+00:00" } }, { "txn_id": "BB-RNW-0218", "assessment_id": "ra_b6451f5f3b6b4e83", "occurred_at": "2026-08-19T02:24:59+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_fa0bbf0eba" }, "output": { "decision": "rate", "score": 0.498, "reason_codes": [], "premium_delta_pct": 5.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.057908+00:00", "fingerprint": "a4ef107adae4845a69f219e9" }, "review": null, "ledger": { "seq": 218, "hash": "98c88576c5eb363e557b052fb010a07a36ff4f687fc17f8ff415702407543857", "prev_hash": "16616119d41ca1247f55e71bfdcbd9e38cc6d69408eaca2df542506b54f61bb9", "recorded_at": "2026-08-19T02:26:57+00:00" } }, { "txn_id": "BB-RNW-0184", "assessment_id": "ra_87b40e38b5ad4431", "occurred_at": "2026-08-19T10:36:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_66494cf9f6" }, "output": { "decision": "rate", "score": 0.414, "reason_codes": [], "premium_delta_pct": 4.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.974948+00:00", "fingerprint": "d70e04efeb7fa23d052cd751" }, "review": null, "ledger": { "seq": 184, "hash": "7f4f08515f618070f088a9caa23fdb4f3667161c88eb082d461932f9e94484c7", "prev_hash": "5351c969b5b00a7e8d978cf50c7464a916ff08ef0cf113660a62cf99d0baef73", "recorded_at": "2026-08-19T10:38:23+00:00" } }, { "txn_id": "BB-RNW-0308", "assessment_id": "ra_f6a3b5341b3048d8", "occurred_at": "2026-08-19T13:39:10+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_11a23a3dda" }, "output": { "decision": "rate", "score": 0.565, "reason_codes": [], "premium_delta_pct": -6.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.292390+00:00", "fingerprint": "2c1e9ee8f3db1166d0e0a439" }, "review": null, "ledger": { "seq": 308, "hash": "ecbdd4077e05474effa864e4b9f9ee15fd5e4813af2a571f1f0213496f38baa8", "prev_hash": "c460870766150e1fa45a12768a0b47912bbf2db80c7f8e72caf2e58f47f24d1f", "recorded_at": "2026-08-19T13:40:49+00:00" } }, { "txn_id": "BB-RNW-0285", "assessment_id": "ra_d01ecead126844fb", "occurred_at": "2026-08-20T21:29:49+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_52b1e1e8b8" }, "output": { "decision": "rate", "score": 0.856, "reason_codes": [], "premium_delta_pct": 10.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-20, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 10 rule(s) in force on 2026-08-20 and recorded 3 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.235959+00:00", "fingerprint": "227c5bede89f91a5c92e3e28" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 10.9, "reviewer_role": "compliance", "resolved_at": "2026-08-21T08:27:12.074057+00:00" }, "ledger": { "seq": 285, "hash": "d274581a833d9ee29406889fd74da3cd9ac7033352cd6d676568131c7af6af04", "prev_hash": "3bb914b962875fe48732ebb6e758008f4f47b5a8c625bdbdf7efdb0d3222608c", "recorded_at": "2026-08-20T21:33:48+00:00" } }, { "txn_id": "BB-RNW-0249", "assessment_id": "ra_d12a13a1b6884098", "occurred_at": "2026-08-21T03:55:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7f488beded" }, "output": { "decision": "rate", "score": 0.872, "reason_codes": [], "premium_delta_pct": 2.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.132862+00:00", "fingerprint": "9e1b804eb4553abcd018f2bb" }, "review": null, "ledger": { "seq": 249, "hash": "117a92988e52a1e2b95ec5e8084d3de1fb144cd40695492a4a60e1404f785906", "prev_hash": "6ff6ed65284b328da4223d885c78ca747d0641fbaf9d0859a85236b2b8fc558b", "recorded_at": "2026-08-21T03:57:55+00:00" } }, { "txn_id": "BB-RNW-0293", "assessment_id": "ra_274921f85a4a4a07", "occurred_at": "2026-08-21T22:19:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_db59f91c4f" }, "output": { "decision": "rate", "score": 0.481, "reason_codes": [], "premium_delta_pct": 10.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-24T22:19:26+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-24 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-21 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.253652+00:00", "fingerprint": "ee4aea8582a6ef1afacbfed6" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 48.5, "reviewer_role": "compliance", "resolved_at": "2026-08-23T22:54:30.432964+00:00" }, "ledger": { "seq": 293, "hash": "a6a2e7d14721c9a7262a1e228b0a4da4c928a145ccb711aae1a4cb202fe25729", "prev_hash": "4f8c51a82c14d866638806b671594d196e19a9e33b085aea6a6db6478d3cb617", "recorded_at": "2026-08-21T22:21:46+00:00" } }, { "txn_id": "BB-RNW-0235", "assessment_id": "ra_1b26468aa2e34ac6", "occurred_at": "2026-08-22T12:08:26+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_225ece4144" }, "output": { "decision": "rate", "score": 0.823, "reason_codes": [], "premium_delta_pct": 21.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-24T12:08:26+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-22, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-24 with specific reasons. RegBench applied 10 rule(s) in force on 2026-08-22 and recorded 3 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.096570+00:00", "fingerprint": "bfd2c0d030266d0a754d142b" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 33.3, "reviewer_role": "compliance", "resolved_at": "2026-08-23T21:29:51.314161+00:00" }, "ledger": { "seq": 235, "hash": "6b4657427af8569f8b637129bf56632ab592253907fc0481cc4f924e1c0f903f", "prev_hash": "5d1791bf99bbc265d012211508b23417e6283b913b933fe741e538bfd0ed73ab", "recorded_at": "2026-08-22T12:12:06+00:00" } }, { "txn_id": "BB-RNW-0219", "assessment_id": "ra_3cf937feac6740ff", "occurred_at": "2026-08-22T12:35:46+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d2303befbd" }, "output": { "decision": "rate", "score": 0.482, "reason_codes": [], "premium_delta_pct": 9.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-22, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.059481+00:00", "fingerprint": "5856bf678e6431e7dab6f2ec" }, "review": null, "ledger": { "seq": 219, "hash": "42df77c129da458198a9c031042cc55b0604fc73fc440417c5a69c7b7b7b1530", "prev_hash": "98c88576c5eb363e557b052fb010a07a36ff4f687fc17f8ff415702407543857", "recorded_at": "2026-08-22T12:38:14+00:00" } }, { "txn_id": "BB-RNW-0221", "assessment_id": "ra_ee8c44d8dbc34c9a", "occurred_at": "2026-08-24T05:49:19+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_63b2835848" }, "output": { "decision": "rate", "score": 0.895, "reason_codes": [], "premium_delta_pct": 20.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-29T05:49:19+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-29 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-24 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.062643+00:00", "fingerprint": "251be38a338b825fe9e611df" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 221, "hash": "7d105a9bda11af1dc380b61e535f73d6da89832ccc2449ef024c7e5551c8eb2f", "prev_hash": "a4a608604c25da7f4771329f9a9c534ae649b14804309351b3d0a44cc2e6de50", "recorded_at": "2026-08-24T05:50:00+00:00" } }, { "txn_id": "BB-RNW-0232", "assessment_id": "ra_759652f5b3cb4c99", "occurred_at": "2026-08-24T11:22:30+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_c4507931f5" }, "output": { "decision": "rate", "score": 0.712, "reason_codes": [], "premium_delta_pct": -5.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-24, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-24 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.085585+00:00", "fingerprint": "01b5eefea56be60553c2aaae" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 8.1, "reviewer_role": "compliance", "resolved_at": "2026-08-24T19:28:07.646070+00:00" }, "ledger": { "seq": 232, "hash": "e6aaf6d74e1c710461d2652016309a716c880ee4ce4f33a308211b18cdcc06c6", "prev_hash": "317c3844fc545bb4b469391d9f21a485d5fd4a467a5a0137a7ef0f57dcdbdfee", "recorded_at": "2026-08-24T11:23:16+00:00" } }, { "txn_id": "BB-RNW-0214", "assessment_id": "ra_586faeed5d104dfb", "occurred_at": "2026-08-25T15:47:39+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_6a285c5e70" }, "output": { "decision": "rate", "score": 0.579, "reason_codes": [], "premium_delta_pct": -3.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.046590+00:00", "fingerprint": "74ad09443c936650faae8b4e" }, "review": null, "ledger": { "seq": 214, "hash": "5f1b640e27ece99b83e1d494309310d48ac03e8ecd7a24f849f47a9e6305775a", "prev_hash": "f962198b8faad803f1ca530a4b2f7de5f26db5c0db601ea8c3a2ca1d67f362fc", "recorded_at": "2026-08-25T15:49:15+00:00" } }, { "txn_id": "BB-RNW-0179", "assessment_id": "ra_a4f35ce087254046", "occurred_at": "2026-08-25T16:56:17+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_26ade941ca" }, "output": { "decision": "rate", "score": 0.784, "reason_codes": [], "premium_delta_pct": 4.4 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.962514+00:00", "fingerprint": "9c9155d533c2a1648a3848ea" }, "review": null, "ledger": { "seq": 179, "hash": "e2a416827ca13d534ae331a835bcb0f274aa269389a731a607265d8dcda961f4", "prev_hash": "f342d0c62aa197ac0450f07bf534f40c83126fa01ef9587a5a1a55c09c0ae74c", "recorded_at": "2026-08-25T16:59:23+00:00" } }, { "txn_id": "BB-RNW-0270", "assessment_id": "ra_13effa1ae45e460d", "occurred_at": "2026-08-25T19:11:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b1431d356b" }, "output": { "decision": "rate", "score": 0.34, "reason_codes": [], "premium_delta_pct": 12.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-29T19:11:31+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-29 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-25 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.185822+00:00", "fingerprint": "ec341fb87e9eaf36eef2172a" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 11.5, "reviewer_role": "compliance", "resolved_at": "2026-08-26T06:41:48.166461+00:00" }, "ledger": { "seq": 270, "hash": "3a9d1ea06784493184c143bece6de9f44e16b5f418ec02050a3fb6b3c5b9f769", "prev_hash": "10276ded90ea43f481a37f16f7ad00350fd12bb46760daf3fa5e6a55398ba40f", "recorded_at": "2026-08-25T19:13:45+00:00" } }, { "txn_id": "BB-RNW-0210", "assessment_id": "ra_375fff5fe5464702", "occurred_at": "2026-08-26T00:39:34+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_442422c2ff" }, "output": { "decision": "rate", "score": 0.386, "reason_codes": [], "premium_delta_pct": -7.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-26, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.035687+00:00", "fingerprint": "2af4238199f35ee2aad72983" }, "review": null, "ledger": { "seq": 210, "hash": "4c42f0945b67d6af10518270e3eda80f3fb2b8dc74e1069c7d7bca62d0cf5a22", "prev_hash": "f890102a9f92d54e9f65c90535c8cdde5a613c8c5b6b7a9badacd6a20d6b6e87", "recorded_at": "2026-08-26T00:41:28+00:00" } }, { "txn_id": "BB-RNW-0236", "assessment_id": "ra_7efd55af246a4a23", "occurred_at": "2026-08-27T10:24:07+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d23909af7c" }, "output": { "decision": "rate", "score": 0.746, "reason_codes": [], "premium_delta_pct": -2.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-27, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-27 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.099462+00:00", "fingerprint": "03b57234b0520018bef1a211" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 10.7, "reviewer_role": "compliance", "resolved_at": "2026-08-27T21:09:21.123760+00:00" }, "ledger": { "seq": 236, "hash": "840ed64e6c8aba3caf714c42de61150610bd70ce8f401f32880b010280e64ba0", "prev_hash": "6b4657427af8569f8b637129bf56632ab592253907fc0481cc4f924e1c0f903f", "recorded_at": "2026-08-27T10:27:07+00:00" } }, { "txn_id": "BB-RNW-0187", "assessment_id": "ra_434d2c511ba24dc3", "occurred_at": "2026-08-28T15:38:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0e8bd6117c" }, "output": { "decision": "rate", "score": 0.862, "reason_codes": [], "premium_delta_pct": 1.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-28, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.979757+00:00", "fingerprint": "4fd55b4791acd12afe7b54c9" }, "review": null, "ledger": { "seq": 187, "hash": "f54f446037ed3847a40ff072c2568fc0def57b422aeb980aa4dc455271bdd05b", "prev_hash": "304436b2e571a3e275c4182e9b9417684bcadfd180c110f9349c935a69d2326a", "recorded_at": "2026-08-28T15:42:00+00:00" } }, { "txn_id": "BB-RNW-0269", "assessment_id": "ra_c407ac2279d143d2", "occurred_at": "2026-08-30T04:10:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_1c937876d8" }, "output": { "decision": "rate", "score": 0.553, "reason_codes": [], "premium_delta_pct": 7.9 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.183636+00:00", "fingerprint": "dba741c7c6f8fddcee0c24d9" }, "review": null, "ledger": { "seq": 269, "hash": "10276ded90ea43f481a37f16f7ad00350fd12bb46760daf3fa5e6a55398ba40f", "prev_hash": "3a5a837d960a6204661d74b1f438127303fb6651172b68347642fde20cedc84b", "recorded_at": "2026-08-30T04:13:11+00:00" } }, { "txn_id": "BB-RNW-0278", "assessment_id": "ra_291e9a071f9a4781", "occurred_at": "2026-08-30T08:13:16+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d9a510ab01" }, "output": { "decision": "rate", "score": 0.548, "reason_codes": [], "premium_delta_pct": 3.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-30, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.211506+00:00", "fingerprint": "3f274e14c7e56ab36b524b94" }, "review": null, "ledger": { "seq": 278, "hash": "1ddd47d8431413bd1b0ead8dcc73dcf1bb28f04c7311d72a9e79abaf75dec153", "prev_hash": "ffac8e1a2e4e388aca9357ed544423d0a8d9453d4a3c3219aee994137f94ccc1", "recorded_at": "2026-08-30T08:16:48+00:00" } }, { "txn_id": "BB-RNW-0318", "assessment_id": "ra_67c625951e0e4829", "occurred_at": "2026-08-31T21:33:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_95d4292eb0" }, "output": { "decision": "rate", "score": 0.398, "reason_codes": [], "premium_delta_pct": -10.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-31, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.315101+00:00", "fingerprint": "206c605408f701cd9ce1bac4" }, "review": null, "ledger": { "seq": 318, "hash": "38b8b6287205c1658f0bff1552c6066df1dd6e94ca51cff9999a9d1aec140bab", "prev_hash": "ac2a6d648ea23bda7ecf101839ded11b2033551a4c0813cd86c3bc9a3c3165a8", "recorded_at": "2026-08-31T21:33:30+00:00" } }, { "txn_id": "BB-RNW-0272", "assessment_id": "ra_a10f89ca278a4a0a", "occurred_at": "2026-09-02T05:01:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_701c82dedd" }, "output": { "decision": "rate", "score": 0.549, "reason_codes": [], "premium_delta_pct": -4.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.193002+00:00", "fingerprint": "efa1c081ec74636c6e788d8c" }, "review": null, "ledger": { "seq": 272, "hash": "79d98ac820a2b4ef30de3c6708cceded119f6807c3f5e080b15d48fc1aa262e5", "prev_hash": "82f57953fa6cddece4221eedc3594579d46a9a5fcd6372cf0f25fb0522fd6698", "recorded_at": "2026-09-02T05:02:09+00:00" } }, { "txn_id": "BB-RNW-0229", "assessment_id": "ra_d79ef281042d494f", "occurred_at": "2026-09-02T10:21:33+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_5774d0abf6" }, "output": { "decision": "rate", "score": 0.544, "reason_codes": [], "premium_delta_pct": -6.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.079114+00:00", "fingerprint": "efa1c081ec74636c6e788d8c" }, "review": null, "ledger": { "seq": 229, "hash": "fa328f2a5aa20a39fefcb69606724ca567ffcfab836aa5b53b2eb4fdae12587d", "prev_hash": "638abc268584b7c9b939af8d3778a4aaa0b631d5090e1b432cf285b869b5295a", "recorded_at": "2026-09-02T10:25:12+00:00" } }, { "txn_id": "BB-RNW-0306", "assessment_id": "ra_6ec8798717fd412a", "occurred_at": "2026-09-02T12:31:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0843722814" }, "output": { "decision": "rate", "score": 0.424, "reason_codes": [], "premium_delta_pct": 6.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.285360+00:00", "fingerprint": "521cc0f0016d8a4dedf9ffc5" }, "review": null, "ledger": { "seq": 306, "hash": "31015601eb96d90e2f53ed26cbfc9d1e996759aa5fcad4da0be914e6f6f92ca2", "prev_hash": "1a09e27c593e14f9791da5c05564a90c01cc5b5e6acabf2d9ff370551471eb0b", "recorded_at": "2026-09-02T12:34:38+00:00" } }, { "txn_id": "BB-RNW-0256", "assessment_id": "ra_c5b8f97107e04812", "occurred_at": "2026-09-02T12:40:56+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7e28af9bea" }, "output": { "decision": "rate", "score": 0.359, "reason_codes": [], "premium_delta_pct": 4.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-02, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-02 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.150700+00:00", "fingerprint": "c5a6cd25ad46be7334050cba" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 14.6, "reviewer_role": "compliance", "resolved_at": "2026-09-03T03:17:47.230003+00:00" }, "ledger": { "seq": 256, "hash": "c20aaf83ebc64ea615faa1a8a781e160bed3fdb3d59bb58ea08296ca9b973838", "prev_hash": "d980903aecc9fd700af508768d5429c09dcda5a532b4cacfa94937f989f31fcb", "recorded_at": "2026-09-02T12:44:17+00:00" } }, { "txn_id": "BB-RNW-0212", "assessment_id": "ra_48c69725a68f4467", "occurred_at": "2026-09-02T16:32:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_882e674caa" }, "output": { "decision": "rate", "score": 0.887, "reason_codes": [], "premium_delta_pct": 12.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-09-02, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-09-02 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.042955+00:00", "fingerprint": "a21d5f1c298d51347ff19d05" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 5.5, "reviewer_role": "compliance", "resolved_at": "2026-09-02T22:05:06.090804+00:00" }, "ledger": { "seq": 212, "hash": "6954cb094ac1a91db2d5db34416494a6378f8dc020f062e6265c1e806a9eab59", "prev_hash": "89898c5ec594534f8353867171a7e46e80845d4b45f908339d23688bb452b2b9", "recorded_at": "2026-09-02T16:34:52+00:00" } }, { "txn_id": "BB-RNW-0158", "assessment_id": "ra_a1ae6e9e2d1c4fb4", "occurred_at": "2026-09-03T00:52:04+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_6518ad80a2" }, "output": { "decision": "rate", "score": 0.75, "reason_codes": [], "premium_delta_pct": 9.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-03, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-03 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.916070+00:00", "fingerprint": "d25ddeb55f859c541bb5b1c4" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 48.9, "reviewer_role": "compliance", "resolved_at": "2026-09-05T01:48:48.221592+00:00" }, "ledger": { "seq": 158, "hash": "26ba249adb81ef572f194c45d8eb17cf9e63083e04131bc07b7566ff348f8995", "prev_hash": "28c03f462f3fe7d9f880d773c5d84743cc7f89e71b0ea7cb683793ec659dd011", "recorded_at": "2026-09-03T00:54:55+00:00" } }, { "txn_id": "BB-RNW-0222", "assessment_id": "ra_e5ed613430fb4854", "occurred_at": "2026-09-03T11:55:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8dd2671c1f" }, "output": { "decision": "rate", "score": 0.31, "reason_codes": [], "premium_delta_pct": 12.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-08T11:55:43+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-08 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-03 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.065045+00:00", "fingerprint": "52780c9c80b6c26709d1ad33" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 19.7, "reviewer_role": "compliance", "resolved_at": "2026-09-04T07:36:48.346138+00:00" }, "ledger": { "seq": 222, "hash": "902a1e4a525726a4998484bc761aa0a9e376ff10ac8e956aeb8e9f008fea9883", "prev_hash": "7d105a9bda11af1dc380b61e535f73d6da89832ccc2449ef024c7e5551c8eb2f", "recorded_at": "2026-09-03T11:57:01+00:00" } }, { "txn_id": "BB-RNW-0205", "assessment_id": "ra_36cf4878264143a7", "occurred_at": "2026-09-04T12:14:55+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e7b652c34f" }, "output": { "decision": "rate", "score": 0.245, "reason_codes": [], "premium_delta_pct": 11.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-08T12:14:55+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-04, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-08 with specific reasons. RegBench applied 10 rule(s) in force on 2026-09-04 and recorded 2 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.024560+00:00", "fingerprint": "8b831f229999b2f5bee1ead4" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 45.1, "reviewer_role": "compliance", "resolved_at": "2026-09-06T09:24:43.362793+00:00" }, "ledger": { "seq": 205, "hash": "8043b12a9f1c6718eb61642368462bdc1ada0e0215912d7edc514b75f9546ccb", "prev_hash": "89f67ef193f5b066e1368de2a19fcbe36c609dc12026f9c5d289232bc5697504", "recorded_at": "2026-09-04T12:17:16+00:00" } }, { "txn_id": "BB-RNW-0230", "assessment_id": "ra_b2ecabe23a694186", "occurred_at": "2026-09-04T14:10:35+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4db1a42b51" }, "output": { "decision": "rate", "score": 0.611, "reason_codes": [], "premium_delta_pct": 2.0 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-04, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-04 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.080932+00:00", "fingerprint": "bbe87a7a36003e698f1851b9" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 28.4, "reviewer_role": "compliance", "resolved_at": "2026-09-05T18:36:38.202277+00:00" }, "ledger": { "seq": 230, "hash": "af7012345dddb2bab57749a916939892ba682d1f3d091d42cf8998e2a4be776e", "prev_hash": "fa328f2a5aa20a39fefcb69606724ca567ffcfab836aa5b53b2eb4fdae12587d", "recorded_at": "2026-09-04T14:13:35+00:00" } }, { "txn_id": "BB-RNW-0258", "assessment_id": "ra_4b1d6b07f70e418e", "occurred_at": "2026-09-04T17:21:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_667bf69194" }, "output": { "decision": "rate", "score": 0.755, "reason_codes": [], "premium_delta_pct": 7.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-04, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.155356+00:00", "fingerprint": "21968297bdcebe00d0eee541" }, "review": null, "ledger": { "seq": 258, "hash": "afe9d190e86677e49ad1cfcb73f538dbe499bb2dae9ac7b603f7206915b9001a", "prev_hash": "ed26cae9e4f8d7276ae83f316c12e6c90eec138bb907e50ea001c3dd071405e6", "recorded_at": "2026-09-04T17:24:32+00:00" } }, { "txn_id": "BB-RNW-0317", "assessment_id": "ra_152f12692935445f", "occurred_at": "2026-09-05T21:38:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_37b0812779" }, "output": { "decision": "rate", "score": 0.209, "reason_codes": [], "premium_delta_pct": 16.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-06T21:38:00+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-05, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-06 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-05 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.313249+00:00", "fingerprint": "beed8d3c4ca98eb1919a8f40" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 28.6, "reviewer_role": "compliance", "resolved_at": "2026-09-07T02:16:45.683749+00:00" }, "ledger": { "seq": 317, "hash": "ac2a6d648ea23bda7ecf101839ded11b2033551a4c0813cd86c3bc9a3c3165a8", "prev_hash": "bf10aede63493421d7ab5e32940cdb9fe2da1a46617fadc82809525f9040db18", "recorded_at": "2026-09-05T21:40:52+00:00" } }, { "txn_id": "BB-RNW-0279", "assessment_id": "ra_012bd7c266b5496e", "occurred_at": "2026-09-06T01:57:45+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_5bf139c5af" }, "output": { "decision": "rate", "score": 0.823, "reason_codes": [], "premium_delta_pct": -0.0 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-06, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-06 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.214653+00:00", "fingerprint": "bbd601b63cff1579a552bf46" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 34.0, "reviewer_role": "compliance", "resolved_at": "2026-09-07T11:58:41.020497+00:00" }, "ledger": { "seq": 279, "hash": "062542f6c06d6ed78bcf178cd0fda9def5b82065d23b16020e3f0e35f71332ec", "prev_hash": "1ddd47d8431413bd1b0ead8dcc73dcf1bb28f04c7311d72a9e79abaf75dec153", "recorded_at": "2026-09-06T02:00:50+00:00" } }, { "txn_id": "BB-RNW-0173", "assessment_id": "ra_7fa3bedaa3a6435b", "occurred_at": "2026-09-07T02:53:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_280072d757" }, "output": { "decision": "rate", "score": 0.884, "reason_codes": [], "premium_delta_pct": 4.3 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.948709+00:00", "fingerprint": "e65c2ee8725e829c36fa3180" }, "review": null, "ledger": { "seq": 173, "hash": "d87ef0f80fdae3fc8917a9381b96df3d416d6111febdd635110e13370f867a82", "prev_hash": "02c9f3954eb9931a94af0e084a608a02ee182ebe46460b6931f83d2ad71d3f88", "recorded_at": "2026-09-07T02:57:25+00:00" } }, { "txn_id": "BB-RNW-0289", "assessment_id": "ra_21945a31d3904fd1", "occurred_at": "2026-09-07T08:33:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_2e0b3c0e6b" }, "output": { "decision": "rate", "score": 0.887, "reason_codes": [], "premium_delta_pct": -8.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.245788+00:00", "fingerprint": "45cd13ede8c144964cef26d4" }, "review": null, "ledger": { "seq": 289, "hash": "bf64ab592a97cc711ea0896682e6cddc8ca38cf8881908b31245785070591617", "prev_hash": "3e1e79b3604aa476f64d12e336aee3589b24e5fbbfa7880eceed91b6f4ac73c8", "recorded_at": "2026-09-07T08:36:00+00:00" } }, { "txn_id": "BB-RNW-0250", "assessment_id": "ra_a809b3614ea44985", "occurred_at": "2026-09-09T00:32:41+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a29eea0d0f" }, "output": { "decision": "rate", "score": 0.626, "reason_codes": [], "premium_delta_pct": 0.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-09, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-09 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.135247+00:00", "fingerprint": "06dbfe4bb5efa7c94e195c66" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 80.7, "reviewer_role": "compliance", "resolved_at": "2026-09-12T09:16:29.293643+00:00" }, "ledger": { "seq": 250, "hash": "f3542fb901ccfa5e79075bf067e77f4398439a8f3c35242cd0e094cd3616447e", "prev_hash": "117a92988e52a1e2b95ec5e8084d3de1fb144cd40695492a4a60e1404f785906", "recorded_at": "2026-09-09T00:34:12+00:00" } }, { "txn_id": "BB-RNW-0290", "assessment_id": "ra_df0f09c87ecb42a5", "occurred_at": "2026-09-11T05:17:18+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_4e7394f7f1" }, "output": { "decision": "rate", "score": 0.539, "reason_codes": [], "premium_delta_pct": -6.8 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-11, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-11 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.247293+00:00", "fingerprint": "02cf0a34a4216e194a5667cb" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 290, "hash": "5faf91d2d7a1249ed77f9e20c871ae155c5400d795bb8a160f8f1a5eae0f9f65", "prev_hash": "bf64ab592a97cc711ea0896682e6cddc8ca38cf8881908b31245785070591617", "recorded_at": "2026-09-11T05:18:38+00:00" } }, { "txn_id": "BB-RNW-0319", "assessment_id": "ra_2e9ddeb0039b492c", "occurred_at": "2026-09-11T15:21:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_91243b3c59" }, "output": { "decision": "rate", "score": 0.602, "reason_codes": [], "premium_delta_pct": 17.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-17T15:21:57+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-11, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-17 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-11 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.316663+00:00", "fingerprint": "dbd3b13bd0714b720ecf1c16" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 12.0, "reviewer_role": "compliance", "resolved_at": "2026-09-12T03:21:55.644827+00:00" }, "ledger": { "seq": 319, "hash": "ca27a2bbacdfb63c9c512822660d92e0c0c8df45f25f097ce6e697b910af0f0d", "prev_hash": "38b8b6287205c1658f0bff1552c6066df1dd6e94ca51cff9999a9d1aec140bab", "recorded_at": "2026-09-11T15:22:49+00:00" } }, { "txn_id": "BB-RNW-0314", "assessment_id": "ra_a55fe61360664357", "occurred_at": "2026-09-13T03:51:43+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_194929e2d7" }, "output": { "decision": "rate", "score": 0.409, "reason_codes": [], "premium_delta_pct": 3.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-13, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.304836+00:00", "fingerprint": "3fbe9738dc7fc2e087fee375" }, "review": null, "ledger": { "seq": 314, "hash": "2696324d8b3a556eb6eecfa74c6b969900a67255f7d48b463ef3e4591e1d7a54", "prev_hash": "256023ea2f2c2a6054c832e52e3a4c5551cec8f4a86502df7e2fc9051ed58482", "recorded_at": "2026-09-13T03:53:07+00:00" } }, { "txn_id": "BB-RNW-0298", "assessment_id": "ra_83d68779297b46d0", "occurred_at": "2026-09-13T08:16:20+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e5a8a78a4b" }, "output": { "decision": "rate", "score": 0.321, "reason_codes": [], "premium_delta_pct": 1.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-13, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-13 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.264751+00:00", "fingerprint": "33c28e0bd243433c035dad95" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 298, "hash": "d28ab0529e37c42731bc7e9d76d5d7fde60c131444f4c602478ce287e4f09cee", "prev_hash": "a5db620b78fb510e4f424a8f5d5f144dc72343e805f3e69f5b5993c933cc2165", "recorded_at": "2026-09-13T08:19:10+00:00" } }, { "txn_id": "BB-RNW-0193", "assessment_id": "ra_3743fd1ebe3f4b75", "occurred_at": "2026-09-13T17:52:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_35457a4e69" }, "output": { "decision": "rate", "score": 0.532, "reason_codes": [], "premium_delta_pct": 6.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-13, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.994492+00:00", "fingerprint": "4ae320210803629c69b208b7" }, "review": null, "ledger": { "seq": 193, "hash": "e3028825e1aa49d37285b4f48884bd2e20dc330bee7ee5c72eb938ab8f218461", "prev_hash": "5677ae0a5560fcfc89709eb7397c6dd0e03bb99ba43053350904afd9b0c450bb", "recorded_at": "2026-09-13T17:55:38+00:00" } }, { "txn_id": "BB-RNW-0267", "assessment_id": "ra_4b62c7c9670e4c6b", "occurred_at": "2026-09-14T13:39:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d3af9444fe" }, "output": { "decision": "rate", "score": 0.38, "reason_codes": [], "premium_delta_pct": 19.2 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-15T13:39:05+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-15 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-14 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.179579+00:00", "fingerprint": "8d5dc21c648ec75c42afd0c2" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 115.3, "reviewer_role": "compliance", "resolved_at": "2026-09-19T09:02:47.618342+00:00" }, "ledger": { "seq": 267, "hash": "50d648af26f6bf26b363c29a7b3765377d55f194a39ec59eb5d9e3777c4f4dec", "prev_hash": "4bdacd8d1f76cc0245d7f3e63544833feeea6793f7ac5598239057f08417a68b", "recorded_at": "2026-09-14T13:43:04+00:00" } }, { "txn_id": "BB-RNW-0283", "assessment_id": "ra_b749b15a090142b1", "occurred_at": "2026-09-14T14:55:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_2eb36b23e3" }, "output": { "decision": "rate", "score": 0.748, "reason_codes": [], "premium_delta_pct": 9.5 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.232608+00:00", "fingerprint": "36e1072503d5ec0fd6e970b7" }, "review": null, "ledger": { "seq": 283, "hash": "22205775bc43b16f120cfcd491976c8f27f6f2c3d32b1b579bf34e1240d5eeff", "prev_hash": "0572d1fe1a2a8b4aa15da33a5629dd5626fc0e1a2dd970d43c8eee880991eb19", "recorded_at": "2026-09-14T14:59:30+00:00" } }, { "txn_id": "BB-RNW-0217", "assessment_id": "ra_84af85a14c254b67", "occurred_at": "2026-09-17T11:35:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_91a72c6d13" }, "output": { "decision": "rate", "score": 0.801, "reason_codes": [], "premium_delta_pct": 18.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-18T11:35:53+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-09-17, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-18 with reasons stated. RegBench applied 11 rule(s) in force on 2026-09-17 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.053617+00:00", "fingerprint": "503912f42b7b2aa51b23a174" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 4.2, "reviewer_role": "compliance", "resolved_at": "2026-09-17T15:53:36.664467+00:00" }, "ledger": { "seq": 217, "hash": "16616119d41ca1247f55e71bfdcbd9e38cc6d69408eaca2df542506b54f61bb9", "prev_hash": "b921178b0743dcce66c17b776e2aab0385445485722df638fbc92f73bf426160", "recorded_at": "2026-09-17T11:38:54+00:00" } }, { "txn_id": "BB-RNW-0259", "assessment_id": "ra_13d6b07505db41be", "occurred_at": "2026-09-18T00:17:29+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_5d0564450a" }, "output": { "decision": "rate", "score": 0.718, "reason_codes": [], "premium_delta_pct": 7.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-18, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-18 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.159040+00:00", "fingerprint": "9cebc7ba5ad88c2f8ce01847" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 54.1, "reviewer_role": "compliance", "resolved_at": "2026-09-20T06:26:15.661434+00:00" }, "ledger": { "seq": 259, "hash": "e409131134b224be2b1e0d2b0c77716dea0759202709e6918831ec16ce956118", "prev_hash": "afe9d190e86677e49ad1cfcb73f538dbe499bb2dae9ac7b603f7206915b9001a", "recorded_at": "2026-09-18T00:18:31+00:00" } }, { "txn_id": "BB-RNW-0226", "assessment_id": "ra_253b0bd2e70249f8", "occurred_at": "2026-09-18T19:09:19+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0f7d020037" }, "output": { "decision": "rate", "score": 0.706, "reason_codes": [], "premium_delta_pct": 15.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-22T19:09:19+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-18, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-22 with specific reasons. RegBench applied 10 rule(s) in force on 2026-09-18 and recorded 2 finding(s); the most serious is 'Adverse outcome from a high-risk AI system with no human review' (medium, L6).", "findings": [ { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.074406+00:00", "fingerprint": "61178379a7b2fe339103f2c7" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 18.7, "reviewer_role": "compliance", "resolved_at": "2026-09-19T13:56:35.556342+00:00" }, "ledger": { "seq": 226, "hash": "fcac7275b64e9ea1b321e185fc552eea1dcc1b71f058ebf40b8b56dce2b72b96", "prev_hash": "8c776e9a1810d4b1af20d8c67da1692daf86c66f931bfdad7bfb0949a1296420", "recorded_at": "2026-09-18T19:12:14+00:00" } }, { "txn_id": "BB-RNW-0225", "assessment_id": "ra_f4dfde589db64a89", "occurred_at": "2026-09-18T20:57:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_026dccfec0" }, "output": { "decision": "rate", "score": 0.334, "reason_codes": [], "premium_delta_pct": -5.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-18, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.070530+00:00", "fingerprint": "658345285f8a1d35bd5e9bf5" }, "review": null, "ledger": { "seq": 225, "hash": "8c776e9a1810d4b1af20d8c67da1692daf86c66f931bfdad7bfb0949a1296420", "prev_hash": "12525bb091e7eb8c65dc246b7a0fd4855bbcbbe30e5cfc6e72b6e2f7c5dcf740", "recorded_at": "2026-09-18T20:59:12+00:00" } }, { "txn_id": "BB-RNW-0213", "assessment_id": "ra_0df31561a5834f17", "occurred_at": "2026-09-20T20:25:05+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_36b4851681" }, "output": { "decision": "rate", "score": 0.352, "reason_codes": [], "premium_delta_pct": -1.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-20, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.045083+00:00", "fingerprint": "3a6a24770bd1e1be6554a53a" }, "review": null, "ledger": { "seq": 213, "hash": "f962198b8faad803f1ca530a4b2f7de5f26db5c0db601ea8c3a2ca1d67f362fc", "prev_hash": "6954cb094ac1a91db2d5db34416494a6378f8dc020f062e6265c1e806a9eab59", "recorded_at": "2026-09-20T20:26:07+00:00" } }, { "txn_id": "BB-RNW-0307", "assessment_id": "ra_b0055d9b42af473e", "occurred_at": "2026-09-21T04:23:18+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_fbdbd6ec84" }, "output": { "decision": "rate", "score": 0.869, "reason_codes": [], "premium_delta_pct": 12.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-25T04:23:18+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-25 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-21 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.286841+00:00", "fingerprint": "9c46f57e9cd3d28f7ef3949e" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 307, "hash": "c460870766150e1fa45a12768a0b47912bbf2db80c7f8e72caf2e58f47f24d1f", "prev_hash": "31015601eb96d90e2f53ed26cbfc9d1e996759aa5fcad4da0be914e6f6f92ca2", "recorded_at": "2026-09-21T04:27:10+00:00" } }, { "txn_id": "BB-RNW-0191", "assessment_id": "ra_0f523856e15c41c1", "occurred_at": "2026-09-21T09:34:17+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_9df58d7dab" }, "output": { "decision": "rate", "score": 0.33, "reason_codes": [], "premium_delta_pct": -1.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-21, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.987876+00:00", "fingerprint": "815f8baa206fb20d8fdc9dda" }, "review": null, "ledger": { "seq": 191, "hash": "6350ae4e8d5eca4a0b5d66f2dee8be38afdd53fce46a2ad6ba2361b961250f57", "prev_hash": "eaf305d9b12d2f0835a4927b00f18f2ecbe14dcdd675c89eb49d8bd88ab4f736", "recorded_at": "2026-09-21T09:34:51+00:00" } }, { "txn_id": "BB-RNW-0297", "assessment_id": "ra_2e2591d9f4854abd", "occurred_at": "2026-09-21T14:46:22+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b35afcc5c9" }, "output": { "decision": "rate", "score": 0.656, "reason_codes": [], "premium_delta_pct": -1.4 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-21, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.263348+00:00", "fingerprint": "6c45779daa94897c51ef673e" }, "review": null, "ledger": { "seq": 297, "hash": "a5db620b78fb510e4f424a8f5d5f144dc72343e805f3e69f5b5993c933cc2165", "prev_hash": "6101712e9a9d58c24db26af66b1ddc75bcd6a572ed6aaceb5cddadbe868746e0", "recorded_at": "2026-09-21T14:49:36+00:00" } }, { "txn_id": "BB-RNW-0264", "assessment_id": "ra_131ad888b2ee4340", "occurred_at": "2026-09-22T21:13:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0fb4140d2a" }, "output": { "decision": "rate", "score": 0.402, "reason_codes": [], "premium_delta_pct": 4.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-22, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.169394+00:00", "fingerprint": "ab1e2a7dadd1b9a1be0e5c44" }, "review": null, "ledger": { "seq": 264, "hash": "05c41cfe3e5e9e86849dc63f4b5f9a9dc4b490af3fc03442aebb3a9e4641b0f9", "prev_hash": "1acc890312f42ba429c28e50d3a224b93ea6a77b60b5480406fc9994208a61f2", "recorded_at": "2026-09-22T21:15:02+00:00" } }, { "txn_id": "BB-RNW-0151", "assessment_id": "ra_f9a3c9bde3694be5", "occurred_at": "2026-09-25T01:02:19+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_16f6d004d0" }, "output": { "decision": "rate", "score": 0.446, "reason_codes": [], "premium_delta_pct": -4.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.900756+00:00", "fingerprint": "81e470083726d5990d0a9651" }, "review": null, "ledger": { "seq": 151, "hash": "46cc2e62b569c6d7f39f893ae9f0ad307d555eee57d02fba95e3deb0978729ac", "prev_hash": "10f72502e0a61807918a646d7543511e2d7d2f0f453a92c311e2950afbad8b55", "recorded_at": "2026-09-25T01:05:09+00:00" } }, { "txn_id": "BB-RNW-0220", "assessment_id": "ra_f9d6205d58fd48fb", "occurred_at": "2026-09-25T16:42:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_c3ecdabc3d" }, "output": { "decision": "rate", "score": 0.507, "reason_codes": [], "premium_delta_pct": -2.7 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.060863+00:00", "fingerprint": "0f3dedb4b99a46f13dec97ab" }, "review": null, "ledger": { "seq": 220, "hash": "a4a608604c25da7f4771329f9a9c534ae649b14804309351b3d0a44cc2e6de50", "prev_hash": "42df77c129da458198a9c031042cc55b0604fc73fc440417c5a69c7b7b7b1530", "recorded_at": "2026-09-25T16:45:20+00:00" } }, { "txn_id": "BB-RNW-0274", "assessment_id": "ra_04f6a185d470432c", "occurred_at": "2026-09-25T17:56:47+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_6da812a0ab" }, "output": { "decision": "rate", "score": 0.499, "reason_codes": [], "premium_delta_pct": -0.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.197388+00:00", "fingerprint": "4891b595e485a3577e47f87e" }, "review": null, "ledger": { "seq": 274, "hash": "f9be1a6e1d066974f7f6dabada4a4f3984b5ec1a26f1819fc1a4714379d28373", "prev_hash": "6e6d9fe1f30eb6b54cab6c9288319476876a569dbf36e02f16a03832ad217bfc", "recorded_at": "2026-09-25T18:00:41+00:00" } }, { "txn_id": "BB-RNW-0185", "assessment_id": "ra_04f8c33195964884", "occurred_at": "2026-09-28T05:45:42+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_062db505be" }, "output": { "decision": "rate", "score": 0.843, "reason_codes": [], "premium_delta_pct": -12.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-28, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.976835+00:00", "fingerprint": "d19d0f7d35478613b5ba270e" }, "review": null, "ledger": { "seq": 185, "hash": "a54ed0e7ae7063ad4eee97bf50bd3eb17993177eed345587a9ef3cb0c6df0e9a", "prev_hash": "7f4f08515f618070f088a9caa23fdb4f3667161c88eb082d461932f9e94484c7", "recorded_at": "2026-09-28T05:48:01+00:00" } }, { "txn_id": "BB-RNW-0216", "assessment_id": "ra_07f1b6a704a6435d", "occurred_at": "2026-09-28T08:45:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_71737cfcee" }, "output": { "decision": "rate", "score": 0.672, "reason_codes": [], "premium_delta_pct": 2.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-28, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.051513+00:00", "fingerprint": "ca63fcb12f172a280108799c" }, "review": null, "ledger": { "seq": 216, "hash": "b921178b0743dcce66c17b776e2aab0385445485722df638fbc92f73bf426160", "prev_hash": "deaca80be5ebc4e9f98b0aab5b8b248061e0cc09e0bbbd480611e9b23077e353", "recorded_at": "2026-09-28T08:46:47+00:00" } }, { "txn_id": "BB-RNW-0200", "assessment_id": "ra_eeb32e32a0124cc9", "occurred_at": "2026-09-30T04:15:15+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7c18f6e749" }, "output": { "decision": "rate", "score": 0.428, "reason_codes": [], "premium_delta_pct": 11.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-06T04:15:15+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-06 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.011116+00:00", "fingerprint": "20b7489b19c864ad1830927a" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 16.7, "reviewer_role": "compliance", "resolved_at": "2026-09-30T21:02:23.060243+00:00" }, "ledger": { "seq": 200, "hash": "5235b613ca27263508f9d34dbdca0f01942a399a20d4dbd5ab75e4a53bfa5a21", "prev_hash": "27bf182dc9ed7236c2be3bab639a4fac7b884d687223b04f9a1f561ca0116716", "recorded_at": "2026-09-30T04:19:08+00:00" } }, { "txn_id": "BB-RNW-0166", "assessment_id": "ra_fbc6af80e38f4ba3", "occurred_at": "2026-09-30T09:12:09+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f086a614b6" }, "output": { "decision": "rate", "score": 0.298, "reason_codes": [], "premium_delta_pct": -6.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-30, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-30 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.934455+00:00", "fingerprint": "b3516c9add8aa1d07e3e452e" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 25.5, "reviewer_role": "compliance", "resolved_at": "2026-10-01T10:46:04.854497+00:00" }, "ledger": { "seq": 166, "hash": "691e97863e85d17572159592d00bf72aaace3fba4a2b5c6ad876126d2a51b4b5", "prev_hash": "937655b5ed3676a40bc371a8f4fa030c3b96d9f0de641f0dd1b768540cdf4062", "recorded_at": "2026-09-30T09:13:24+00:00" } }, { "txn_id": "BB-RNW-0276", "assessment_id": "ra_72b903a8f2ac4e01", "occurred_at": "2026-09-30T09:59:55+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b090e7d048" }, "output": { "decision": "rate", "score": 0.778, "reason_codes": [], "premium_delta_pct": 7.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.202516+00:00", "fingerprint": "eb101043ddd96fff430a7169" }, "review": null, "ledger": { "seq": 276, "hash": "0ed499a75c9a79b7342a4c776a4bbe9e0b4eeea1cfc23abeb98f98a33c0eed55", "prev_hash": "52194654e203ece5b151a293a5d75ce8681f6e6fe2ee89e8dcb8a806810ca879", "recorded_at": "2026-09-30T10:03:08+00:00" } }, { "txn_id": "BB-RNW-0240", "assessment_id": "ra_20ea0716d44b4406", "occurred_at": "2026-09-30T12:45:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_fa1319cd0e" }, "output": { "decision": "rate", "score": 0.685, "reason_codes": [], "premium_delta_pct": 19.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-06T12:45:57+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-06 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.110631+00:00", "fingerprint": "45652e2600fea5e11492430d" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 35.1, "reviewer_role": "compliance", "resolved_at": "2026-10-01T23:55:01.168574+00:00" }, "ledger": { "seq": 240, "hash": "3e8334236a9df5ea5295cc8f600a8af9ecd5ef403a631bb371275302300f2db3", "prev_hash": "d8cd5e908cf7f2b37d9e72e36955a155a164cefaf6b8dd589000711ffffef1a7", "recorded_at": "2026-09-30T12:46:46+00:00" } }, { "txn_id": "BB-RNW-0302", "assessment_id": "ra_8c43b1977bf045d0", "occurred_at": "2026-09-30T18:11:07+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_78bad46f84" }, "output": { "decision": "rate", "score": 0.21, "reason_codes": [], "premium_delta_pct": 7.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.272066+00:00", "fingerprint": "92d411075fd44501a4459085" }, "review": null, "ledger": { "seq": 302, "hash": "94e868cd051acba0a1f492e349ba8e872b781cc10e5ce26464c9dcf03a48d780", "prev_hash": "a68a9338f5050f0bd8b3b8de992d738c0576ed8c6558c8d76a1117470ac58689", "recorded_at": "2026-09-30T18:14:24+00:00" } }, { "txn_id": "BB-RNW-0171", "assessment_id": "ra_caf74521b9e346e4", "occurred_at": "2026-10-01T03:31:06+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f51295fbb6" }, "output": { "decision": "rate", "score": 0.365, "reason_codes": [], "premium_delta_pct": 13.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-06T03:31:06+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-01, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-06 with specific reasons. RegBench applied 10 rule(s) in force on 2026-10-01 and recorded 3 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.944932+00:00", "fingerprint": "a897ee4b3a603e985aae8312" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 30.3, "reviewer_role": "compliance", "resolved_at": "2026-10-02T09:49:51.594711+00:00" }, "ledger": { "seq": 171, "hash": "9e5133e7975b481c6b4e330fcb0ff677f2cff5914a7836b8a9a105affebc01df", "prev_hash": "547da407234726de9aca5bf0504836237ad57be1984896e37632833391843323", "recorded_at": "2026-10-01T03:32:43+00:00" } }, { "txn_id": "BB-RNW-0239", "assessment_id": "ra_fc5ae227163c433b", "occurred_at": "2026-10-01T03:39:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_df484e9059" }, "output": { "decision": "rate", "score": 0.428, "reason_codes": [], "premium_delta_pct": -0.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-01, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-01 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-01", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.108522+00:00", "fingerprint": "6599e93a86828571ca1e2820" }, "review": null, "ledger": { "seq": 239, "hash": "d8cd5e908cf7f2b37d9e72e36955a155a164cefaf6b8dd589000711ffffef1a7", "prev_hash": "24159b593b9597fba47ea1db4217fe04d7cf27450d6e6a951f3eb38628f82118", "recorded_at": "2026-10-01T03:43:10+00:00" } }, { "txn_id": "BB-RNW-0223", "assessment_id": "ra_b30f124a197a4e11", "occurred_at": "2026-10-03T10:24:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8460f823f6" }, "output": { "decision": "rate", "score": 0.611, "reason_codes": [], "premium_delta_pct": -7.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-03, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.066994+00:00", "fingerprint": "495c038be6d0e96d98befcae" }, "review": null, "ledger": { "seq": 223, "hash": "3b12b3d50be82afa5bad692185dfdc12cac98f6b0b792ea3849e0317e52172cd", "prev_hash": "902a1e4a525726a4998484bc761aa0a9e376ff10ac8e956aeb8e9f008fea9883", "recorded_at": "2026-10-03T10:28:40+00:00" } }, { "txn_id": "BB-RNW-0224", "assessment_id": "ra_f00733e1b1f14c09", "occurred_at": "2026-10-04T03:21:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_759809fe21" }, "output": { "decision": "rate", "score": 0.831, "reason_codes": [], "premium_delta_pct": 3.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-04, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.068621+00:00", "fingerprint": "41c003cadd0556b3cf1eae39" }, "review": null, "ledger": { "seq": 224, "hash": "12525bb091e7eb8c65dc246b7a0fd4855bbcbbe30e5cfc6e72b6e2f7c5dcf740", "prev_hash": "3b12b3d50be82afa5bad692185dfdc12cac98f6b0b792ea3849e0317e52172cd", "recorded_at": "2026-10-04T03:22:58+00:00" } }, { "txn_id": "BB-RNW-0180", "assessment_id": "ra_8a44c7a77fb6479b", "occurred_at": "2026-10-04T04:37:19+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_414f780c0c" }, "output": { "decision": "rate", "score": 0.481, "reason_codes": [], "premium_delta_pct": -5.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-04, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-04 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.965860+00:00", "fingerprint": "0108d405d271ab4f37b865b8" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 60.0, "reviewer_role": "compliance", "resolved_at": "2026-10-06T16:36:51.624794+00:00" }, "ledger": { "seq": 180, "hash": "a82086851eb714053de9d49db92d090727e2c514f9b36921053e3e6ce365a259", "prev_hash": "e2a416827ca13d534ae331a835bcb0f274aa269389a731a607265d8dcda961f4", "recorded_at": "2026-10-04T04:39:27+00:00" } }, { "txn_id": "BB-RNW-0155", "assessment_id": "ra_5550839f952d4442", "occurred_at": "2026-10-04T09:32:20+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d60fb8c0f7" }, "output": { "decision": "rate", "score": 0.559, "reason_codes": [], "premium_delta_pct": -13.8 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-04, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.909279+00:00", "fingerprint": "23c3fc518b972cd8a4f429f9" }, "review": null, "ledger": { "seq": 155, "hash": "f1cb04da97a9acca33bab9e9b9d72a537307fb5327053e1417dd48d357f9c67c", "prev_hash": "281b3781cf051db7b74b67007cd07fa66f58b47ea4d0d5c9d396bcd138a5c8f9", "recorded_at": "2026-10-04T09:34:01+00:00" } }, { "txn_id": "BB-RNW-0189", "assessment_id": "ra_09ee8c94ab3f4e0f", "occurred_at": "2026-10-05T15:52:10+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_1c16a3d845" }, "output": { "decision": "rate", "score": 0.75, "reason_codes": [], "premium_delta_pct": 5.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-05, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.983086+00:00", "fingerprint": "bd2d1b1dedf42514d2d313d3" }, "review": null, "ledger": { "seq": 189, "hash": "ecf9616b240805bd58c797dd0afbc410989bfd03fbd76516539b81355defc7e5", "prev_hash": "3a2d67d4f66284b30eaaaa853b30b550b2ecb0aa32727040a5222c8d4c7445ad", "recorded_at": "2026-10-05T15:55:07+00:00" } }, { "txn_id": "BB-RNW-0316", "assessment_id": "ra_9262bccaa6ac479d", "occurred_at": "2026-10-05T19:57:10+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_03f165b930" }, "output": { "decision": "rate", "score": 0.751, "reason_codes": [], "premium_delta_pct": 0.5 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-05, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-05 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.310927+00:00", "fingerprint": "a5b93688662912dcc2f89085" }, "review": null, "ledger": { "seq": 316, "hash": "bf10aede63493421d7ab5e32940cdb9fe2da1a46617fadc82809525f9040db18", "prev_hash": "887d27ee022645df00ea56b974b5abb32107bb26b7ad37039680cf50c9ca6cd3", "recorded_at": "2026-10-05T19:59:51+00:00" } }, { "txn_id": "BB-RNW-0284", "assessment_id": "ra_3d6156db08d342ac", "occurred_at": "2026-10-07T02:39:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7eba7b17b7" }, "output": { "decision": "rate", "score": 0.718, "reason_codes": [], "premium_delta_pct": 16.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-12T02:39:03+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-12 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-07 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.234340+00:00", "fingerprint": "f38824c6b450d27a6720b550" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 12.7, "reviewer_role": "compliance", "resolved_at": "2026-10-07T15:26:26.684170+00:00" }, "ledger": { "seq": 284, "hash": "3bb914b962875fe48732ebb6e758008f4f47b5a8c625bdbdf7efdb0d3222608c", "prev_hash": "22205775bc43b16f120cfcd491976c8f27f6f2c3d32b1b579bf34e1240d5eeff", "recorded_at": "2026-10-07T02:42:35+00:00" } }, { "txn_id": "BB-RNW-0299", "assessment_id": "ra_21e695b5405a4e51", "occurred_at": "2026-10-07T09:38:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8ac0b3dfc1" }, "output": { "decision": "rate", "score": 0.891, "reason_codes": [], "premium_delta_pct": 6.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.266353+00:00", "fingerprint": "78a3ce3356100907fff45c4b" }, "review": null, "ledger": { "seq": 299, "hash": "6c33a8289bd638c8ac96d96ca7fd96ab13480e8d28d52738a233b823818863e7", "prev_hash": "d28ab0529e37c42731bc7e9d76d5d7fde60c131444f4c602478ce287e4f09cee", "recorded_at": "2026-10-07T09:39:32+00:00" } }, { "txn_id": "BB-RNW-0190", "assessment_id": "ra_a8cf89e0f6924aa1", "occurred_at": "2026-10-07T22:07:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_20cfef16c3" }, "output": { "decision": "rate", "score": 0.209, "reason_codes": [], "premium_delta_pct": -6.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.985015+00:00", "fingerprint": "fc348ac7613e1ebbdeda539e" }, "review": null, "ledger": { "seq": 190, "hash": "eaf305d9b12d2f0835a4927b00f18f2ecbe14dcdd675c89eb49d8bd88ab4f736", "prev_hash": "ecf9616b240805bd58c797dd0afbc410989bfd03fbd76516539b81355defc7e5", "recorded_at": "2026-10-07T22:09:58+00:00" } }, { "txn_id": "CO-AUTO-0144", "assessment_id": "ra_43edcf2152f04f8e", "occurred_at": "2026-10-09T15:00:00+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "telematics://bluebonnet/program", "cra://demo-bureau/score" ], "consumer_token": "tok_showcase_0144" }, "output": { "decision": "rate", "score": 0.72, "reason_codes": [], "premium_delta_pct": 14.5 }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-11T15:00:00+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": "Colorado auto renewal priced with telematics and credit-based score; the same transaction dated before 15 October 2025 falls outside Regulation 10-1-1's amended scope.", "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-09, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-11 with specific reasons. RegBench applied 10 rule(s) in force on 2026-10-09 and recorded 3 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:25.363251+00:00", "fingerprint": "9b1dd85f202155207e64dfc2" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 6.8, "reviewer_role": "compliance", "resolved_at": "2026-10-09T21:52:57.224387+00:00" }, "ledger": { "seq": 1125, "hash": "852d7d9347da922a27f5afce10ff456ee67fe02425498f4426a8456d1b223d44", "prev_hash": "dba2ecd14e3ab92057b32340d5c7a1dcb8a74ac3c1f828915a3a2db979d42971", "recorded_at": "2026-10-09T15:02:01+00:00" } }, { "txn_id": "BB-RNW-0309", "assessment_id": "ra_8547924977ee45c5", "occurred_at": "2026-10-13T00:04:57+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_bf383b5f46" }, "output": { "decision": "rate", "score": 0.354, "reason_codes": [], "premium_delta_pct": 20.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-18T00:04:57+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-13, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-18 with specific reasons. RegBench applied 10 rule(s) in force on 2026-10-13 and recorded 2 finding(s); the most serious is 'Adverse outcome from a high-risk AI system with no human review' (medium, L6).", "findings": [ { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.294800+00:00", "fingerprint": "3eed413347a60749bb500336" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 19.7, "reviewer_role": "compliance", "resolved_at": "2026-10-13T19:48:36.908869+00:00" }, "ledger": { "seq": 309, "hash": "e520e7f243ec61ed3bdf23f7336ad0a07f3b5c697a54d79cbf5be155bddd6939", "prev_hash": "ecbdd4077e05474effa864e4b9f9ee15fd5e4813af2a571f1f0213496f38baa8", "recorded_at": "2026-10-13T00:06:53+00:00" } }, { "txn_id": "BB-RNW-0242", "assessment_id": "ra_ea7c2fd01e9b4f5b", "occurred_at": "2026-10-14T02:01:07+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_5b3c3ccd8b" }, "output": { "decision": "rate", "score": 0.723, "reason_codes": [], "premium_delta_pct": -9.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.114176+00:00", "fingerprint": "1973262ed56d442cfbd5ba4a" }, "review": null, "ledger": { "seq": 242, "hash": "97f22c1330b5fc0bf0bffd6e7deafd323eb82db0f9d1fa9734c57011b829da6c", "prev_hash": "a087a56849041ec9c9e181c8dea1806d4e5f956cdc636183d5708393583cf56a", "recorded_at": "2026-10-14T02:02:28+00:00" } }, { "txn_id": "BB-RNW-0172", "assessment_id": "ra_845d223915854c1e", "occurred_at": "2026-10-14T21:35:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_c5041433c0" }, "output": { "decision": "rate", "score": 0.378, "reason_codes": [], "premium_delta_pct": 21.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-16T21:35:26+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-14, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-16 with reasons stated. RegBench applied 11 rule(s) in force on 2026-10-14 and recorded 2 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.946742+00:00", "fingerprint": "4486133130a9d5055f17adce" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 8.1, "reviewer_role": "compliance", "resolved_at": "2026-10-15T05:42:45.010017+00:00" }, "ledger": { "seq": 172, "hash": "02c9f3954eb9931a94af0e084a608a02ee182ebe46460b6931f83d2ad71d3f88", "prev_hash": "9e5133e7975b481c6b4e330fcb0ff677f2cff5914a7836b8a9a105affebc01df", "recorded_at": "2026-10-14T21:38:18+00:00" } }, { "txn_id": "BB-RNW-0320", "assessment_id": "ra_0c51cc95865a49d2", "occurred_at": "2026-10-18T01:03:17+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_851da5b96c" }, "output": { "decision": "rate", "score": 0.431, "reason_codes": [], "premium_delta_pct": -8.9 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-18, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.319193+00:00", "fingerprint": "9a461d64157ae41e0eadb9a0" }, "review": null, "ledger": { "seq": 320, "hash": "f1da339dbaf4cf1db70e52e97d15a36d9aec700e2eb787650a93b202442758c2", "prev_hash": "ca27a2bbacdfb63c9c512822660d92e0c0c8df45f25f097ce6e697b910af0f0d", "recorded_at": "2026-10-18T01:04:44+00:00" } }, { "txn_id": "BB-RNW-0233", "assessment_id": "ra_71160b0031794b62", "occurred_at": "2026-10-20T15:56:44+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7f354e9239" }, "output": { "decision": "rate", "score": 0.761, "reason_codes": [], "premium_delta_pct": -4.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-20, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.092232+00:00", "fingerprint": "8a214bd0648648c076a0aa2e" }, "review": null, "ledger": { "seq": 233, "hash": "1e7d10fea127bad4abaf023cd419ea20155ecbbc1557c49ff67d751d91ac693d", "prev_hash": "e6aaf6d74e1c710461d2652016309a716c880ee4ce4f33a308211b18cdcc06c6", "recorded_at": "2026-10-20T16:00:42+00:00" } }, { "txn_id": "BB-RNW-0286", "assessment_id": "ra_a2b19af9e1254ab6", "occurred_at": "2026-10-21T00:01:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_6c703ab6f1" }, "output": { "decision": "rate", "score": 0.39, "reason_codes": [], "premium_delta_pct": 15.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-25T00:01:00+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-25 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-21 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.237707+00:00", "fingerprint": "439b1ba04a7601c594c43123" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 31.4, "reviewer_role": "compliance", "resolved_at": "2026-10-22T07:30:53.213960+00:00" }, "ledger": { "seq": 286, "hash": "2bbd305a21832e7093ab51f62fd6476cbc49bb28623e498e8caaa5b091a9a1c5", "prev_hash": "d274581a833d9ee29406889fd74da3cd9ac7033352cd6d676568131c7af6af04", "recorded_at": "2026-10-21T00:04:20+00:00" } }, { "txn_id": "BB-RNW-0265", "assessment_id": "ra_601e2bff434f4a36", "occurred_at": "2026-10-21T00:46:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f56b1afd2b" }, "output": { "decision": "rate", "score": 0.856, "reason_codes": [], "premium_delta_pct": 13.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-24T00:46:32+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-24 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-21 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.174081+00:00", "fingerprint": "ee0b8738aa3c460e3b578bc8" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 46.6, "reviewer_role": "compliance", "resolved_at": "2026-10-22T23:28:22.519932+00:00" }, "ledger": { "seq": 265, "hash": "14304034db2f312bc140c3694a96d641d04d36405b8583cf2f30bc654ba2df5f", "prev_hash": "05c41cfe3e5e9e86849dc63f4b5f9a9dc4b490af3fc03442aebb3a9e4641b0f9", "recorded_at": "2026-10-21T00:50:10+00:00" } }, { "txn_id": "BB-RNW-0300", "assessment_id": "ra_dc3b994c35044d56", "occurred_at": "2026-10-21T10:27:50+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_89238c5d30" }, "output": { "decision": "rate", "score": 0.226, "reason_codes": [], "premium_delta_pct": 7.1 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.268700+00:00", "fingerprint": "0d23041f71ed6fd95d6b14d5" }, "review": null, "ledger": { "seq": 300, "hash": "117c6e195d92f81e25d55de9c942316b3cf988dc844c2b6731090418dbf597bc", "prev_hash": "6c33a8289bd638c8ac96d96ca7fd96ab13480e8d28d52738a233b823818863e7", "recorded_at": "2026-10-21T10:28:51+00:00" } }, { "txn_id": "BB-RNW-0169", "assessment_id": "ra_9790cdc536514cb7", "occurred_at": "2026-10-21T14:52:43+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_996481a0ff" }, "output": { "decision": "rate", "score": 0.544, "reason_codes": [], "premium_delta_pct": 5.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-21 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.941137+00:00", "fingerprint": "a7a0f57cf9cc1b1d0edf0b72" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 51.1, "reviewer_role": "compliance", "resolved_at": "2026-10-23T18:02:34.129400+00:00" }, "ledger": { "seq": 169, "hash": "9fc2b501ba4c44b73e9f15fb74a3624810d98aac201ed8e3c9084589eebf55b1", "prev_hash": "8cfe1e1f5308589b0a35c7100cd25592e701846f521b8c64d4a7a2bb119d00c3", "recorded_at": "2026-10-21T14:56:33+00:00" } }, { "txn_id": "BB-RNW-0254", "assessment_id": "ra_b81f4ddc96cc4571", "occurred_at": "2026-10-21T20:42:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_8ece7f0fe7" }, "output": { "decision": "rate", "score": 0.697, "reason_codes": [], "premium_delta_pct": -1.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.147202+00:00", "fingerprint": "8df97af6bbec67226c2942e3" }, "review": null, "ledger": { "seq": 254, "hash": "fb51237e0434437f02b124eed32546740bf64f811d3d23b798e8a41514a50df8", "prev_hash": "bc248fac7c707679aed58ae5b8d5853becf0e6e0d919a6d304e6a3a167776ff2", "recorded_at": "2026-10-21T20:46:43+00:00" } }, { "txn_id": "BB-RNW-0159", "assessment_id": "ra_0b525fa66d9244f8", "occurred_at": "2026-10-21T23:15:38+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_3276be0529" }, "output": { "decision": "rate", "score": 0.283, "reason_codes": [], "premium_delta_pct": 3.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-21, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-21 and recorded 1 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.918991+00:00", "fingerprint": "73382e234f2ce2b159c1a9a0" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 32.8, "reviewer_role": "compliance", "resolved_at": "2026-10-23T08:05:40.820135+00:00" }, "ledger": { "seq": 159, "hash": "3827e70cb8010bf5cadb068138e7515c98628586bb24c6b63c7ff495d7c2554e", "prev_hash": "26ba249adb81ef572f194c45d8eb17cf9e63083e04131bc07b7566ff348f8995", "recorded_at": "2026-10-21T23:18:46+00:00" } }, { "txn_id": "BB-RNW-0231", "assessment_id": "ra_663a060a7cc74b57", "occurred_at": "2026-10-22T02:14:53+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_93e4c66d4e" }, "output": { "decision": "rate", "score": 0.899, "reason_codes": [], "premium_delta_pct": -7.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.083275+00:00", "fingerprint": "3265f682724ad51a74065978" }, "review": null, "ledger": { "seq": 231, "hash": "317c3844fc545bb4b469391d9f21a485d5fd4a467a5a0137a7ef0f57dcdbdfee", "prev_hash": "af7012345dddb2bab57749a916939892ba682d1f3d091d42cf8998e2a4be776e", "recorded_at": "2026-10-22T02:18:14+00:00" } }, { "txn_id": "BB-RNW-0275", "assessment_id": "ra_f49f9674aa1a44d2", "occurred_at": "2026-10-24T05:59:30+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_7ba819f7ac" }, "output": { "decision": "rate", "score": 0.597, "reason_codes": [], "premium_delta_pct": 10.5 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-27T05:59:30+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-24, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-27 with reasons stated. RegBench applied 10 rule(s) in force on 2026-10-24 and recorded 2 finding(s); the most serious is 'ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)' (medium, L3).", "findings": [ { "rule_id": "co.reg-10-1-1.governance-framework-amended", "layer": "L3", "judgment_basis": "black_letter", "advisory": false, "severity": "medium", "title": "ECDIS used outside an attested governance and risk management framework (amended Regulation 10-1-1)", "required_action": "Add the system to the Regulation 10-1-1 inventory, attest framework coverage, and include it in the next annual report filed through SERFF.", "exhibits": [ "A", "B" ], "citations": [ { "provision_id": "co-reg-10-1-1-s5", "label": "3 CCR 702-10, Regulation 10-1-1 (amended, effective 15 Oct 2025), \u00a7 5.A \u2014 governance and risk management framework (life, private passenger auto, health benefit plans)", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true }, { "provision_id": "co-reg-10-1-1-s6", "label": "3 CCR 702-10, Regulation 10-1-1 (amended), \u00a7 6 \u2014 reporting requirements and officer attestation", "source_url": "https://www.sos.state.co.us/CCR/GenerateRulePdf.do?ruleVersionId=12182", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.199257+00:00", "fingerprint": "6e04e0b79479871637509960" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 11.5, "reviewer_role": "compliance", "resolved_at": "2026-10-24T17:33:48.555228+00:00" }, "ledger": { "seq": 275, "hash": "52194654e203ece5b151a293a5d75ce8681f6e6fe2ee89e8dcb8a806810ca879", "prev_hash": "f9be1a6e1d066974f7f6dabada4a4f3984b5ec1a26f1819fc1a4714379d28373", "recorded_at": "2026-10-24T06:02:25+00:00" } }, { "txn_id": "BB-RNW-0281", "assessment_id": "ra_f08f4325b7154912", "occurred_at": "2026-10-24T09:49:59+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0be050085b" }, "output": { "decision": "rate", "score": 0.515, "reason_codes": [], "premium_delta_pct": 19.2 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-26T09:49:59+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": true, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-26 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-24 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.225805+00:00", "fingerprint": "651710ad6add6b97ac19d730" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 24.2, "reviewer_role": "compliance", "resolved_at": "2026-10-25T10:03:23.577331+00:00" }, "ledger": { "seq": 281, "hash": "b05f367c94addc5a889703e4b45dc923bfcc32785f9eae2d5aa035af5b10b488", "prev_hash": "46bbc154043ac8a51d8163cec5c297125590623295044c26b5544fb9d11c0ba6", "recorded_at": "2026-10-24T09:51:49+00:00" } }, { "txn_id": "BB-RNW-0263", "assessment_id": "ra_47ac308f9df747d5", "occurred_at": "2026-10-24T11:16:35+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_f073fc32a6" }, "output": { "decision": "rate", "score": 0.666, "reason_codes": [], "premium_delta_pct": -6.7 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-24, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.166782+00:00", "fingerprint": "c157a706132c7e24a0900aa1" }, "review": null, "ledger": { "seq": 263, "hash": "1acc890312f42ba429c28e50d3a224b93ea6a77b60b5480406fc9994208a61f2", "prev_hash": "64ee02850f7ea2b6cdb2455d4f1c82f4671f6c0bc5f45d24635639e180fae2c3", "recorded_at": "2026-10-24T11:17:54+00:00" } }, { "txn_id": "BB-RNW-0262", "assessment_id": "ra_4eb88b3e696e4bf4", "occurred_at": "2026-10-24T15:37:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_e160d9e44b" }, "output": { "decision": "rate", "score": 0.469, "reason_codes": [], "premium_delta_pct": 5.8 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-24, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.163979+00:00", "fingerprint": "41cd7ed52cd2aefa0b4b61ec" }, "review": null, "ledger": { "seq": 262, "hash": "64ee02850f7ea2b6cdb2455d4f1c82f4671f6c0bc5f45d24635639e180fae2c3", "prev_hash": "c1dbe2646796c411319e148fb38a3b09878bc7157e2813b83f727df5436fde9b", "recorded_at": "2026-10-24T15:39:22+00:00" } }, { "txn_id": "BB-RNW-0282", "assessment_id": "ra_cb71cc9797f045c6", "occurred_at": "2026-10-25T02:05:49+00:00", "state": "CO", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_93ab1ca59a" }, "output": { "decision": "rate", "score": 0.784, "reason_codes": [], "premium_delta_pct": 8.7 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-25, price-opt v7.2 (an internally developed system) produced 'rate' in a CO auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. External consumer data and information sources were present. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.229289+00:00", "fingerprint": "3bc36108edb132dcaefcafd4" }, "review": null, "ledger": { "seq": 282, "hash": "0572d1fe1a2a8b4aa15da33a5629dd5626fc0e1a2dd970d43c8eee880991eb19", "prev_hash": "b05f367c94addc5a889703e4b45dc923bfcc32785f9eae2d5aa035af5b10b488", "recorded_at": "2026-10-25T02:07:19+00:00" } }, { "txn_id": "BB-RNW-0183", "assessment_id": "ra_8e335afb766d4627", "occurred_at": "2026-10-25T19:40:44+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_632aac2048" }, "output": { "decision": "rate", "score": 0.637, "reason_codes": [], "premium_delta_pct": -0.3 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-25, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.971464+00:00", "fingerprint": "e3a9fe131a4169dbcaab19f1" }, "review": null, "ledger": { "seq": 183, "hash": "5351c969b5b00a7e8d978cf50c7464a916ff08ef0cf113660a62cf99d0baef73", "prev_hash": "211e058d2055d05c3f77f226164200c17323313240015437d8462b301151699b", "recorded_at": "2026-10-25T19:41:51+00:00" } }, { "txn_id": "BB-RNW-0195", "assessment_id": "ra_6e4cd3c443d94829", "occurred_at": "2026-10-26T07:40:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_a8fa69ac2c" }, "output": { "decision": "rate", "score": 0.321, "reason_codes": [], "premium_delta_pct": 9.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.997968+00:00", "fingerprint": "0398dca774549b3c2b063bd8" }, "review": null, "ledger": { "seq": 195, "hash": "eaea6173a4f94a44436a6b5c91e85d0952921dae9f48cc2f78c4e5c27f2a181b", "prev_hash": "a580c3c55cd0aa94aa6aaed1031a1178addf50938ac4b74a8f1b306a006bfec0", "recorded_at": "2026-10-26T07:43:04+00:00" } }, { "txn_id": "BB-RNW-0211", "assessment_id": "ra_367b11b01d864521", "occurred_at": "2026-10-26T18:01:08+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_0b805efa36" }, "output": { "decision": "rate", "score": 0.506, "reason_codes": [], "premium_delta_pct": -3.4 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.037826+00:00", "fingerprint": "b77c322dff54b1facb178cd4" }, "review": null, "ledger": { "seq": 211, "hash": "89898c5ec594534f8353867171a7e46e80845d4b45f908339d23688bb452b2b9", "prev_hash": "4c42f0945b67d6af10518270e3eda80f3fb2b8dc74e1069c7d7bca62d0cf5a22", "recorded_at": "2026-10-26T18:02:44+00:00" } }, { "txn_id": "BB-RNW-0301", "assessment_id": "ra_c66a15bfbe684449", "occurred_at": "2026-10-27T09:17:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_ff2f7501c6" }, "output": { "decision": "rate", "score": 0.394, "reason_codes": [], "premium_delta_pct": 6.5 }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-27, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.270152+00:00", "fingerprint": "b40fbfc19014154a947172d8" }, "review": null, "ledger": { "seq": 301, "hash": "a68a9338f5050f0bd8b3b8de992d738c0576ed8c6558c8d76a1117470ac58689", "prev_hash": "117c6e195d92f81e25d55de9c942316b3cf988dc844c2b6731090418dbf597bc", "recorded_at": "2026-10-27T09:18:29+00:00" } }, { "txn_id": "BB-RNW-0207", "assessment_id": "ra_589abb3eab89479c", "occurred_at": "2026-10-27T12:14:20+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_d6ce1c807e" }, "output": { "decision": "rate", "score": 0.649, "reason_codes": [], "premium_delta_pct": 2.1 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-27, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.029293+00:00", "fingerprint": "4ea87c2a56e14b2b76028cb9" }, "review": null, "ledger": { "seq": 207, "hash": "790f76f32c1e37c4e55a50ea2f3436230f65ae29b1a9ec9cde3aea008bba0f66", "prev_hash": "8199733c16df8c6f0b6f787c0ba7167a27c5c159b7467c0e0b7587243aab0e83", "recorded_at": "2026-10-27T12:16:25+00:00" } }, { "txn_id": "BB-RNW-0304", "assessment_id": "ra_dfaa479321d3466c", "occurred_at": "2026-10-27T17:36:12+00:00", "state": "TX", "line_of_business": "auto", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "price-opt", "model_version": "7.2", "vendor": "internal", "vendor_name": null, "high_risk_class": "pricing" }, "inputs": { "feature_classes": [ "telematics", "credit", "geography", "vehicle", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/score", "telematics://bluebonnet/program" ], "consumer_token": "tok_b3d912f298" }, "output": { "decision": "rate", "score": 0.237, "reason_codes": [], "premium_delta_pct": 8.6 }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-27, price-opt v7.2 (an internally developed system) produced 'rate' in a TX auto renewal transaction. Inputs included telematics, credit, geography, vehicle, claims history. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:23.278630+00:00", "fingerprint": "f3951565d0ca2ce2560552dd" }, "review": null, "ledger": { "seq": 304, "hash": "c139c7d282672f418fb5999aa3faba2d6c7909d8e613f6f9b657deed50cf52f8", "prev_hash": "bd3d3558e57ec617747839ec4cf22a22b3795a5fec0d784e5068c83ee4a6535c", "recorded_at": "2026-10-27T17:39:32+00:00" } } ] }, { "system_id": "uw-triage", "items": { "1_name_version": "Underwriting Triage \u2014 versions observed: 2.4 (declared 2.4)", "2_use_cases": "Scores new auto and home applications for eligibility and refers borderline risks to an underwriter. Stages: {'underwriting': 149, 'renewal': 1}; decisions in period: 150; adverse: 26; states: {'TX': 115, 'CO': 35}.", "3_model_type": "GLM", "3a_agentic": { "agentic": false, "objectives": null, "authorizations_without_human_approval": null }, "4_implementation_date": "2024-03-01", "5_development": { "vendor": "internal", "vendor_name": null }, "6_purchased_modifiable": { "purchased": false, "developed_for_company": true, "modifiable": true }, "7_model_risks": { "description": "Potential adverse consumer impact evidenced by findings on the model's decisions in the period.", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "severity": "medium", "layer": "L6", "count": 25 }, { "rule_id": "us.fcra.permissible-purpose-undocumented", "title": "Consumer report used with no data source reference recorded", "severity": "low", "layer": "L0", "count": 14 }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "title": "Third-party data used with no source recorded", "severity": "low", "layer": "L4", "count": 13 }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "title": "Adverse outcome from a high-risk AI system with no human review", "severity": "medium", "layer": "L6", "count": 11 }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "severity": "high", "layer": "L4", "count": 10 }, { "rule_id": "tx.559.credit-adverse-effect-notice", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "severity": "high", "layer": "L2", "count": 6 }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "severity": "medium", "layer": "L5", "count": 5 }, { "rule_id": "us.fcra.adverse-action-notice", "title": "Adverse action based on a consumer report with no FCRA notice recorded", "severity": "critical", "layer": "L0", "count": 4 } ], "adverse_decisions": 26, "dispositions": { "FLAG": 33, "PASS": 111, "HOLD": 6 } }, "8_limitations": { "carrier_documentation": "no model documentation reference recorded on the decision stream", "regbench_note": "RegBench assesses recorded facts about each decision; it does not inspect model internals or detect inputs outside the training domain." }, "9_automate_augment_support": { "declared": "augment", "observed_human_review_pct": 38.7 }, "10_testing_validation": { "testing_report_ref_on_decisions": [ "test://bluebonnet/uw-triage/2026Q3" ], "regbench_continuous_assessment": "150 of 150 decisions assessed at decision time against the rules in force on each decision date." }, "10a_proxy_evaluation": [ { "system_id": "uw-triage", "model_version": "2.4", "tenant_id": "bluebonnet", "state": null, "stage_mix": { "underwriting": 149, "renewal": 1 }, "reference_group": "A", "groups": [ { "group": "A", "decisions": 83, "adverse": 13, "adverse_rate": 0.1566 } ], "suppressed_groups": [ "C", "B", "D" ], "flags": [], "method": "Adverse-outcome rate by carrier-estimated demographic proxy group; reference group = largest group; adverse impact ratio = group rate / reference rate (flag below 0.80); marginal difference in percentage points (flag at 5.0 or more). Cells under 30 decisions are suppressed. Illustrative thresholds.", "computed_where": "carrier environment (proxy labels never leave it); only these aggregates are released" } ], "11_last_test_date": { "carrier": "see testing report reference", "regbench_last_assessment": "2026-10-27T19:21:41+00:00" }, "12_financial_statement_effect": "material", "13_legal_compliance_review": { "rules_applied": [ "doctrine.unreviewed-adverse-high-risk", "doctrine.zip-credit-interaction-proxy", "precedent.ai-flagged-cancellation-without-notice", "tx.559.credit-adverse-effect-notice", "tx.tdi.consequential-decision-without-human-review", "tx.tdi.third-party-data-source-unrecorded", "us.fcra.adverse-action-notice", "us.fcra.permissible-purpose-undocumented" ], "provisions_cited": [ "naic-ai-bulletin-s3-1-ais-program", "precedent-pa-geico-2026", "tx-ins-559-001", "tx-ins-559-054", "tx-tdi-b-0003-26-governance", "tx-tdi-b-0003-26-human-review", "tx-tdi-b-0036-20", "us-fcra-1681a-k-1-b-i", "us-fcra-1681b-a-3-c", "us-fcra-1681m-a" ], "method": "Deterministic rule evaluation per decision; citations resolve against the corpus snapshot in force on the decision date or the assessment fails closed." }, "14_regulatory_actions": "None recorded in RegBench. Carrier to confirm any informal agreements, administrative complaints, consent orders or investigations." }, "case_files": [ { "txn_id": "BB-UW-0045", "assessment_id": "ra_0ec97aa06c0a4d75", "occurred_at": "2026-07-06T17:32:22+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_502dfbc7b2" }, "output": { "decision": "decline", "score": 0.844, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-07T17:32:22+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-06, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-07 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-06 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.633058+00:00", "fingerprint": "0b17eaf391b21f3654ba9b4e" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 9.6, "reviewer_role": "underwriter", "resolved_at": "2026-07-07T03:13:58.122743+00:00" }, "ledger": { "seq": 45, "hash": "8a9cdc4431d7c83b488b1de7997065a4b3d3b59f98715a5fed6f736764f1bf44", "prev_hash": "a4d8638aec9c14b6f5dffa688655402843599c3396806fda6fa8e4bf72c1988b", "recorded_at": "2026-07-06T17:35:52+00:00" } }, { "txn_id": "BB-UW-0123", "assessment_id": "ra_b9f3b76384df4fed", "occurred_at": "2026-07-07T00:39:02+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_1ae2f4c4c9" }, "output": { "decision": "approve", "score": 0.437, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-07, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.842820+00:00", "fingerprint": "8ab662803c5ed703dbb80ba8" }, "review": null, "ledger": { "seq": 123, "hash": "2ca80e3f8da0a367dc3dd87750c3fb5ed42d5ccfa9476ff3484454910e60a48d", "prev_hash": "7d2652fb9e2398d4c1709224d5dcf7430d20a970c0c15a0d8801924470bf7cfd", "recorded_at": "2026-07-07T00:41:05+00:00" } }, { "txn_id": "BB-UW-0099", "assessment_id": "ra_fcd1c47239f34f2a", "occurred_at": "2026-07-08T06:31:12+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9fbc24bd96" }, "output": { "decision": "decline", "score": 0.888, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-08, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-07-08 and recorded 4 finding(s); the most serious is 'Adverse action based on a consumer report with no FCRA notice recorded' (high, L0).", "findings": [ { "rule_id": "us.fcra.adverse-action-notice", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Adverse action based on a consumer report with no FCRA notice recorded", "required_action": "Hold the outcome for compliance review; generate the notice with every required element before the decision is communicated, and record notice_sent_at on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "us-fcra-1681m-a", "label": "15 U.S.C. \u00a7 1681m(a) (FCRA \u00a7 615(a)) \u2014 duties of users taking adverse actions on the basis of a consumer report", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681m", "verified": true }, { "provision_id": "us-fcra-1681a-k-1-b-i", "label": "15 U.S.C. \u00a7 1681a(k)(1)(B)(i) \u2014 'adverse action' in connection with insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681a", "verified": true } ] }, { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "layer": "L5", "judgment_basis": "precedent", "advisory": true, "severity": "medium", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "required_action": "Confirm the notice and document deadlines before the cancellation or decline takes effect; record the notice on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "precedent-pa-geico-2026", "label": "Pennsylvania Attorney General agreement with GEICO (announced May 2026) \u2014 AI-enabled underwriting review and cancellation without adequate notice", "source_url": "https://www.attorneygeneral.gov/?p=33518", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.767207+00:00", "fingerprint": "374b97a68978d9a6298518e9" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 5.0, "reviewer_role": "underwriter", "resolved_at": "2026-07-08T11:31:20.369162+00:00" }, "ledger": { "seq": 99, "hash": "989887f541d517f6a05da1f110b0788af8887394fec52609f8793fd59add5cab", "prev_hash": "bf3ac86a5a03e810e044519e0b49dbbf2d921c951a43e43bb41f6cd778709547", "recorded_at": "2026-07-08T06:34:12+00:00" } }, { "txn_id": "BB-UW-0040", "assessment_id": "ra_92ebf327b1d241f4", "occurred_at": "2026-07-08T06:47:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2ca39d2086" }, "output": { "decision": "approve", "score": 0.105, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-08, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.623848+00:00", "fingerprint": "d692f39c446d5507667729f9" }, "review": null, "ledger": { "seq": 40, "hash": "a336e1644abcf3d96613124bb4ead1587d958291082af6599ae774200473cd5f", "prev_hash": "99935d8b892761762b7e6c539631abf1c29b506166abf72fe4df3072200ca525", "recorded_at": "2026-07-08T06:50:30+00:00" } }, { "txn_id": "BB-UW-0146", "assessment_id": "ra_40c2dcbfbf4048b8", "occurred_at": "2026-07-09T08:04:23+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_50a054ddcf" }, "output": { "decision": "approve", "score": 0.358, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-07-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.892280+00:00", "fingerprint": "23e7021060e5566e7c75ba6d" }, "review": null, "ledger": { "seq": 146, "hash": "62574275b17d245129eb50350d11ab61d10aca1776cd964141fe10138a86622e", "prev_hash": "08dde01e7a1d2cceb4224246854858505666aae56b707f7cbc795fa96a3bbdaa", "recorded_at": "2026-07-09T08:05:14+00:00" } }, { "txn_id": "BB-UW-0020", "assessment_id": "ra_da67a67505db4852", "occurred_at": "2026-07-10T08:34:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_de73aa881e" }, "output": { "decision": "approve", "score": 0.401, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.574767+00:00", "fingerprint": "cec144c18373ec4d11834c25" }, "review": null, "ledger": { "seq": 20, "hash": "bb4c30cb67dda82960cad0dab6c491c10c1b7e86c3c9da2c920ff91d874128aa", "prev_hash": "a4f4f78a3658f13cc57deb539d01ded356b28c0e4165830fbcfcbb79a73511bd", "recorded_at": "2026-07-10T08:35:15+00:00" } }, { "txn_id": "BB-UW-0150", "assessment_id": "ra_2f37b2ebf5a3433b", "occurred_at": "2026-07-10T16:00:39+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_4ed980f95d" }, "output": { "decision": "refer", "score": 0.527, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-10, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.899331+00:00", "fingerprint": "2e170be35aeb8b2cb1b49ed7" }, "review": null, "ledger": { "seq": 150, "hash": "10f72502e0a61807918a646d7543511e2d7d2f0f453a92c311e2950afbad8b55", "prev_hash": "df811ec2ede6a80779192bb96005ae0c3661a5ece2b5e1ca92cc42f10a0a8640", "recorded_at": "2026-07-10T16:03:45+00:00" } }, { "txn_id": "BB-UW-0017", "assessment_id": "ra_fd2105b0da2f4576", "occurred_at": "2026-07-10T23:25:35+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_58d573e7bb" }, "output": { "decision": "refer", "score": 0.456, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-10, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-10 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.565503+00:00", "fingerprint": "2e170be35aeb8b2cb1b49ed7" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 20.6, "reviewer_role": "underwriter", "resolved_at": "2026-07-11T20:01:25.847080+00:00" }, "ledger": { "seq": 17, "hash": "3a3dd6b1fbfd7b19ddd0d74b21758584f8116ba76a1146a814c0d8bef3bdf13d", "prev_hash": "f51a072b17b97b8ebca62fe8457825f3e9c558237af7b3ae32c5c1aee547d55d", "recorded_at": "2026-07-10T23:27:35+00:00" } }, { "txn_id": "BB-UW-0037", "assessment_id": "ra_7c72e05c4a9a4ea6", "occurred_at": "2026-07-12T07:40:59+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_38f85e9bea" }, "output": { "decision": "decline", "score": 0.954, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-12, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-07-12 and recorded 4 finding(s); the most serious is 'Adverse action based on a consumer report with no FCRA notice recorded' (high, L0).", "findings": [ { "rule_id": "us.fcra.adverse-action-notice", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Adverse action based on a consumer report with no FCRA notice recorded", "required_action": "Hold the outcome for compliance review; generate the notice with every required element before the decision is communicated, and record notice_sent_at on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "us-fcra-1681m-a", "label": "15 U.S.C. \u00a7 1681m(a) (FCRA \u00a7 615(a)) \u2014 duties of users taking adverse actions on the basis of a consumer report", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681m", "verified": true }, { "provision_id": "us-fcra-1681a-k-1-b-i", "label": "15 U.S.C. \u00a7 1681a(k)(1)(B)(i) \u2014 'adverse action' in connection with insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681a", "verified": true } ] }, { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "layer": "L5", "judgment_basis": "precedent", "advisory": true, "severity": "medium", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "required_action": "Confirm the notice and document deadlines before the cancellation or decline takes effect; record the notice on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "precedent-pa-geico-2026", "label": "Pennsylvania Attorney General agreement with GEICO (announced May 2026) \u2014 AI-enabled underwriting review and cancellation without adequate notice", "source_url": "https://www.attorneygeneral.gov/?p=33518", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.615151+00:00", "fingerprint": "591b438ca5e281a69d0dcb2a" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 14.5, "reviewer_role": "underwriter", "resolved_at": "2026-07-12T22:15:06.987299+00:00" }, "ledger": { "seq": 37, "hash": "80fcb2a9e7b3d862345ce98ffea91f5f9138235672a783c1aea474f345fcdc3e", "prev_hash": "2660771fe0ef1edbe35350bb3af7d8cd69fa5eddb5ed8565a2846595848af2b2", "recorded_at": "2026-07-12T07:43:05+00:00" } }, { "txn_id": "BB-UW-0121", "assessment_id": "ra_e91af32a614e4d7e", "occurred_at": "2026-07-14T07:17:02+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_f7741152c4" }, "output": { "decision": "approve", "score": 0.304, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.837813+00:00", "fingerprint": "c327f77abca848b2e203042e" }, "review": null, "ledger": { "seq": 121, "hash": "a5f82666abb03898593d82022a71294a234c4db3a7f33a4fcef52532c85b87af", "prev_hash": "9ccc5cb2a91ee75121dbd3d32410f28772f03128ff9743e75b1d9542602b1d37", "recorded_at": "2026-07-14T07:20:22+00:00" } }, { "txn_id": "BB-UW-0078", "assessment_id": "ra_78c19e72de7f4b91", "occurred_at": "2026-07-14T19:28:04+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9ac0a904a0" }, "output": { "decision": "decline", "score": 0.794, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-07-14, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded 4 finding(s); the most serious is 'Adverse action based on a consumer report with no FCRA notice recorded' (high, L0).", "findings": [ { "rule_id": "us.fcra.adverse-action-notice", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Adverse action based on a consumer report with no FCRA notice recorded", "required_action": "Hold the outcome for compliance review; generate the notice with every required element before the decision is communicated, and record notice_sent_at on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "us-fcra-1681m-a", "label": "15 U.S.C. \u00a7 1681m(a) (FCRA \u00a7 615(a)) \u2014 duties of users taking adverse actions on the basis of a consumer report", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681m", "verified": true }, { "provision_id": "us-fcra-1681a-k-1-b-i", "label": "15 U.S.C. \u00a7 1681a(k)(1)(B)(i) \u2014 'adverse action' in connection with insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681a", "verified": true } ] }, { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "layer": "L5", "judgment_basis": "precedent", "advisory": true, "severity": "medium", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "required_action": "Confirm the notice and document deadlines before the cancellation or decline takes effect; record the notice on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "precedent-pa-geico-2026", "label": "Pennsylvania Attorney General agreement with GEICO (announced May 2026) \u2014 AI-enabled underwriting review and cancellation without adequate notice", "source_url": "https://www.attorneygeneral.gov/?p=33518", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.703050+00:00", "fingerprint": "8da775081ca1693a1812fb1f" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 5.0, "reviewer_role": "underwriter", "resolved_at": "2026-07-15T00:30:08.061117+00:00" }, "ledger": { "seq": 78, "hash": "ebc3d5fb4ced408c0b88c4d0a9ee2f1ae5d374f56eda56113c7c72b11b8ad4ad", "prev_hash": "0f5090f4f191b7cef2a61574f4002d057e71a5ab106cce40c57a48c71846e945", "recorded_at": "2026-07-14T19:28:37+00:00" } }, { "txn_id": "BB-UW-0086", "assessment_id": "ra_f457a86fb14b485a", "occurred_at": "2026-07-14T20:20:05+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_b3e2b770ee" }, "output": { "decision": "approve", "score": 0.283, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.733185+00:00", "fingerprint": "5366940e3a2462cb7f58a467" }, "review": null, "ledger": { "seq": 86, "hash": "79c778bbe13efc31b9d0c5eda792c1acb3216efd04e0f76c717d27adf3f4d4c1", "prev_hash": "524170528f96cafdd4e1ea20a0f1ea2d8e23192957980234039f1d00ba539e70", "recorded_at": "2026-07-14T20:23:58+00:00" } }, { "txn_id": "BB-UW-0129", "assessment_id": "ra_48cc9975db8f4dc7", "occurred_at": "2026-07-14T22:48:21+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_507a90d4df" }, "output": { "decision": "refer", "score": 0.678, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-14, uw-triage v2.4 (an internally developed system) produced 'refer' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-07-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.854275+00:00", "fingerprint": "f828789ea8cc8f007316c435" }, "review": null, "ledger": { "seq": 129, "hash": "0b94a501a6a47c795540fd4c603f3006d38d5ccfca1446028255c842045c30d8", "prev_hash": "63c2a0bbabfb55f92050df6655cf38f72006a707007179d2ccafadaead7b1a83", "recorded_at": "2026-07-14T22:51:31+00:00" } }, { "txn_id": "BB-UW-0089", "assessment_id": "ra_28ff56b51e604f8a", "occurred_at": "2026-07-15T10:20:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_c0bf71ca17" }, "output": { "decision": "decline", "score": 0.918, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-21T10:20:04+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 5 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-15, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-07-21 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-15 and recorded 5 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.745775+00:00", "fingerprint": "11f8a4af3fadb7f47999fbae" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 89, "hash": "aec84993463018fc2067c0a1f3742b8f38f0d49f83d3387c024c6e80456ea6c6", "prev_hash": "eeb0502df7c4baf4c662d87f5a1a90acf8c0c20d2ad7a6a15982f33d56cfdf2f", "recorded_at": "2026-07-15T10:22:31+00:00" } }, { "txn_id": "BB-UW-0149", "assessment_id": "ra_6b03ff686ed449ac", "occurred_at": "2026-07-16T01:42:47+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_1ae5eab899" }, "output": { "decision": "approve", "score": 0.052, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-16, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-07-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.897936+00:00", "fingerprint": "8d66fa5878e56737ad7a40b4" }, "review": null, "ledger": { "seq": 149, "hash": "df811ec2ede6a80779192bb96005ae0c3661a5ece2b5e1ca92cc42f10a0a8640", "prev_hash": "ed5e042b94915834033dc792ce5b1b77be3d5885a8bdc5b36edc1c1e7cad6a2b", "recorded_at": "2026-07-16T01:43:27+00:00" } }, { "txn_id": "BB-UW-0075", "assessment_id": "ra_fe6bc6b9136e4be0", "occurred_at": "2026-07-16T07:21:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7776035237" }, "output": { "decision": "approve", "score": 0.262, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-16, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.697773+00:00", "fingerprint": "e3e3ae1a525882e44a817fa9" }, "review": null, "ledger": { "seq": 75, "hash": "d9e66d632853bfc5c62aa6fede7fe03538999d9fd75eaac686656a86a5238c44", "prev_hash": "f53de172bd571eccdbb2354d5aa30be25f7305da54dead054baad58898e22133", "recorded_at": "2026-07-16T07:23:58+00:00" } }, { "txn_id": "BB-UW-0048", "assessment_id": "ra_97fd6c7fff80487a", "occurred_at": "2026-07-17T00:42:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_c711ac44be" }, "output": { "decision": "approve", "score": 0.156, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-17, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.641519+00:00", "fingerprint": "7047054c674aa67c79f7ce92" }, "review": null, "ledger": { "seq": 48, "hash": "260aa859521fc49eddcfcb70a85518622bc5d300058c0e85b76980081c667f6e", "prev_hash": "2337016d96de052d6e34ea7b576dd02359c7c5e3be956cfe74a600829c904ccd", "recorded_at": "2026-07-17T00:45:40+00:00" } }, { "txn_id": "BB-UW-0110", "assessment_id": "ra_7e26071a04144439", "occurred_at": "2026-07-17T01:45:07+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_c493ef1690" }, "output": { "decision": "decline", "score": 0.943, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-07-22T01:45:07+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-17, uw-triage v2.4 (an internally developed system) produced 'decline' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-07-22 with specific reasons. RegBench applied 10 rule(s) in force on 2026-07-17 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.812307+00:00", "fingerprint": "c4f7a6e4c464747edb7cd7ba" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 33.0, "reviewer_role": "underwriter", "resolved_at": "2026-07-18T10:45:20.713556+00:00" }, "ledger": { "seq": 110, "hash": "4c8c3a648814b180765c3b35c9f788df11a53afc3fc079d2477fb2bcdbb374b2", "prev_hash": "24633efa909dfd177e3a043b086462d39c7f07836a4f85631c8d56f1b69c7e69", "recorded_at": "2026-07-17T01:46:01+00:00" } }, { "txn_id": "BB-UW-0013", "assessment_id": "ra_7504ba5e019b492b", "occurred_at": "2026-07-17T03:06:48+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_cb9c31b930" }, "output": { "decision": "approve", "score": 0.269, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-17, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.551652+00:00", "fingerprint": "ea69d361b56f97cb36cc5698" }, "review": null, "ledger": { "seq": 13, "hash": "30d81e2f6f18978c45615271ba590f54d2efd77646c821f5ad642e15ddf790b6", "prev_hash": "ce337d78c19c99fe66216ee1999520fc41b7be9a999b03efaf3f77fd9a36857a", "recorded_at": "2026-07-17T03:08:19+00:00" } }, { "txn_id": "BB-UW-0124", "assessment_id": "ra_dc268a8c364646f7", "occurred_at": "2026-07-19T00:04:33+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_8d46c569ce" }, "output": { "decision": "approve", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.844754+00:00", "fingerprint": "b678b3acc4316ce72e954490" }, "review": null, "ledger": { "seq": 124, "hash": "d6042b00a34bf863e11a98f8dab1c57fe5a5b16442bc1d0afa403329c8ba7975", "prev_hash": "2ca80e3f8da0a367dc3dd87750c3fb5ed42d5ccfa9476ff3484454910e60a48d", "recorded_at": "2026-07-19T00:06:49+00:00" } }, { "txn_id": "BB-UW-0145", "assessment_id": "ra_f0984f12e59548ac", "occurred_at": "2026-07-19T00:40:15+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5e9e3db699" }, "output": { "decision": "approve", "score": 0.274, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.888619+00:00", "fingerprint": "b678b3acc4316ce72e954490" }, "review": null, "ledger": { "seq": 145, "hash": "08dde01e7a1d2cceb4224246854858505666aae56b707f7cbc795fa96a3bbdaa", "prev_hash": "7f41bb1cecbcea5c283820b3f327f2ea1646461f1b7fe1fedcf2d5d2b9a68689", "recorded_at": "2026-07-19T00:43:50+00:00" } }, { "txn_id": "BB-UW-0131", "assessment_id": "ra_1e4aa64dce41468d", "occurred_at": "2026-07-19T05:02:46+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7f40c5f5d2" }, "output": { "decision": "approve", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.860205+00:00", "fingerprint": "1274a40c2489ae8ba0c4084e" }, "review": null, "ledger": { "seq": 131, "hash": "44d826bbfce7ae2d012744c6a08455775c2ba4a4ce9d94185282de7ec42357d4", "prev_hash": "35f833b7b4f5c05095e0886e5deb6d52e71513055efce85d5eac7128b6964c6f", "recorded_at": "2026-07-19T05:04:36+00:00" } }, { "txn_id": "BB-UW-0027", "assessment_id": "ra_d6314f3a30dd403f", "occurred_at": "2026-07-21T15:29:14+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5081b2499e" }, "output": { "decision": "approve", "score": 0.176, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.589754+00:00", "fingerprint": "d28a678e6c4b435de6620a97" }, "review": null, "ledger": { "seq": 27, "hash": "5afe4da57af98212fe9ba24eae9b1521be5296853307ae2dd766c570347b1eed", "prev_hash": "8761189c2723dc846aa47e7904f92a10ed6392ba4f034fbd579ad628c218dc02", "recorded_at": "2026-07-21T15:30:21+00:00" } }, { "txn_id": "BB-UW-0118", "assessment_id": "ra_97ab42c7461942cf", "occurred_at": "2026-07-21T23:45:09+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6bdfc9f54d" }, "output": { "decision": "approve", "score": 0.092, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-21, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.833524+00:00", "fingerprint": "2587d7b944d62a9bd3914fc6" }, "review": null, "ledger": { "seq": 118, "hash": "c11caa235cff8ade9f6a735859d9d81b168d4ef7b721de4e5b1a2d5440c6d650", "prev_hash": "0095f5b056445b0f9565b39b0b8a2b7c9901ca326822b70983ddc45669d06955", "recorded_at": "2026-07-21T23:47:27+00:00" } }, { "txn_id": "BB-UW-0112", "assessment_id": "ra_3d2d7c51a0c945ed", "occurred_at": "2026-07-22T10:08:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_9bdf4a5fe5" }, "output": { "decision": "approve", "score": 0.185, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-22 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.817930+00:00", "fingerprint": "46104f868d2bbc6571ea881d" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 11.6, "reviewer_role": "underwriter", "resolved_at": "2026-07-22T21:46:17.237346+00:00" }, "ledger": { "seq": 112, "hash": "798964458d908cf91683265745a15e3dbd3f683901ee7060ce3c5dbd1ccf3c6a", "prev_hash": "9580e44fba902edc18378abc355beaf8168625777dff1814948d90446b26497e", "recorded_at": "2026-07-22T10:11:36+00:00" } }, { "txn_id": "BB-UW-0093", "assessment_id": "ra_44566a6294ef4ecc", "occurred_at": "2026-07-22T23:14:37+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_201148da69" }, "output": { "decision": "approve", "score": 0.426, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.757385+00:00", "fingerprint": "6eb9ddd4bf7dabd89e7e4776" }, "review": null, "ledger": { "seq": 93, "hash": "6105350864a44d85b3dc0fc5e5e4992ac706ac3866b89fd96fd76b0cf84574ed", "prev_hash": "ff637bbd3491ccaafa060fb937eaeff5c811ef97d05368f6a7820d2e4735324f", "recorded_at": "2026-07-22T23:17:17+00:00" } }, { "txn_id": "BB-UW-0047", "assessment_id": "ra_300bea88784d46c2", "occurred_at": "2026-07-23T19:49:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7959095c12" }, "output": { "decision": "approve", "score": 0.381, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.639450+00:00", "fingerprint": "e030122d23c171fa822d85ea" }, "review": null, "ledger": { "seq": 47, "hash": "2337016d96de052d6e34ea7b576dd02359c7c5e3be956cfe74a600829c904ccd", "prev_hash": "e3332ee401cddcc0f865ef0e8eeaca28ebfa72b533746392ccb54b0a22f11951", "recorded_at": "2026-07-23T19:49:56+00:00" } }, { "txn_id": "BB-UW-0142", "assessment_id": "ra_8a1517c25f9d4af8", "occurred_at": "2026-07-24T05:40:58+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_afb2f1ee98" }, "output": { "decision": "approve", "score": 0.102, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-24, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.883385+00:00", "fingerprint": "f2d867ff0f554aebd5d63eb5" }, "review": null, "ledger": { "seq": 142, "hash": "0bbda96f8baca277ecbb1e915e816cdd02c80c0fd80e7b6c501b8673601ce30f", "prev_hash": "5dfd89edecd1e6838c827c6e3954e754e8ffc834aac2f49a6d293ae3bb795ae7", "recorded_at": "2026-07-24T05:42:20+00:00" } }, { "txn_id": "BB-UW-0023", "assessment_id": "ra_dba46d19824d4cb8", "occurred_at": "2026-07-24T07:43:57+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_cb86f9fa21" }, "output": { "decision": "approve", "score": 0.136, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-24, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.580418+00:00", "fingerprint": "33644dcf1f9c3c0ca5600cd4" }, "review": null, "ledger": { "seq": 23, "hash": "3e1f6a418069dab68648a6fd83dfbfcc152dd674142231388d34061cd89740ca", "prev_hash": "17ad1ed4b2b876f89e1475030cc9cc4ebf6a8ae31affff609f37edfacb0224b6", "recorded_at": "2026-07-24T07:44:57+00:00" } }, { "txn_id": "BB-UW-0136", "assessment_id": "ra_cff370af6fe24f22", "occurred_at": "2026-07-24T13:06:14+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_bd09d471f0" }, "output": { "decision": "approve", "score": 0.411, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-24, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.869116+00:00", "fingerprint": "30854f1f661e917f41eb507e" }, "review": null, "ledger": { "seq": 136, "hash": "4e61ecffcd51833d52b9b7a2c79a04e30d3d10419093a9a1876034b980e9c1fa", "prev_hash": "cfa7b35b29970665018de771f5f8fb93895ad3f4f45e61bf7a10acdef27becab", "recorded_at": "2026-07-24T13:09:57+00:00" } }, { "txn_id": "BB-UW-0122", "assessment_id": "ra_0838cfde876449ea", "occurred_at": "2026-07-26T00:09:28+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_316e043667" }, "output": { "decision": "approve", "score": 0.11, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.841146+00:00", "fingerprint": "709c60b8d8ced1be8d1561ac" }, "review": null, "ledger": { "seq": 122, "hash": "7d2652fb9e2398d4c1709224d5dcf7430d20a970c0c15a0d8801924470bf7cfd", "prev_hash": "a5f82666abb03898593d82022a71294a234c4db3a7f33a4fcef52532c85b87af", "recorded_at": "2026-07-26T00:13:15+00:00" } }, { "txn_id": "BB-UW-0054", "assessment_id": "ra_0f3c51d45a6e4fca", "occurred_at": "2026-07-26T08:38:10+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_945581ebc2" }, "output": { "decision": "approve", "score": 0.24, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.651695+00:00", "fingerprint": "278da6d4269759109cdc1287" }, "review": null, "ledger": { "seq": 54, "hash": "ab3a80cfa4885abdf192fd4ee63efa30a80f0231668351cfd9967821b5778514", "prev_hash": "27c407513956bb9bd8c14393eb4a6592d191f2df1d5dca580c4eabadeb55e2c6", "recorded_at": "2026-07-26T08:39:37+00:00" } }, { "txn_id": "BB-UW-0090", "assessment_id": "ra_97aeb17a26ae436f", "occurred_at": "2026-07-27T22:19:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_d00062c542" }, "output": { "decision": "approve", "score": 0.217, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-27, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-27 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.749488+00:00", "fingerprint": "ba7049d0923e26e5d48655e9" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 10.8, "reviewer_role": "underwriter", "resolved_at": "2026-07-28T09:07:49.560574+00:00" }, "ledger": { "seq": 90, "hash": "2a2d9076d74636ee742243252f24775dbce40cf65d0808a885bf9210d1cc3639", "prev_hash": "aec84993463018fc2067c0a1f3742b8f38f0d49f83d3387c024c6e80456ea6c6", "recorded_at": "2026-07-27T22:22:22+00:00" } }, { "txn_id": "BB-UW-0065", "assessment_id": "ra_decea15e15d94bf1", "occurred_at": "2026-07-28T22:59:10+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_cc219a09b4" }, "output": { "decision": "approve", "score": 0.317, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-28, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.675843+00:00", "fingerprint": "82ccb990509db421c4befac7" }, "review": null, "ledger": { "seq": 65, "hash": "4bc10a5aa51dd7fb3d8a0d10cfb20bf046331a198866824f8ee16966329ea5a9", "prev_hash": "7062a019656d4180156b841993d8b18d7e5af371e59638eb60c110a2ea816f76", "recorded_at": "2026-07-28T23:02:50+00:00" } }, { "txn_id": "BB-UW-0105", "assessment_id": "ra_a6fba96cf4ce4c49", "occurred_at": "2026-07-30T06:13:23+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a075ff71f6" }, "output": { "decision": "decline", "score": 0.753, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-02T06:13:23+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-07-30, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-02 with specific reasons. RegBench applied 11 rule(s) in force on 2026-07-30 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-07-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.792858+00:00", "fingerprint": "67416916ff2dae193575c02e" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 17.6, "reviewer_role": "underwriter", "resolved_at": "2026-07-30T23:52:42.037376+00:00" }, "ledger": { "seq": 105, "hash": "668edaa09a4050c60b565728bf161710328c4ddfe4c2a0d3a00c41bab5605fe8", "prev_hash": "1adf359a02c5e05aa106eabb00c16c0023f9ea8562951ef5de3c40f49b325bfd", "recorded_at": "2026-07-30T06:15:02+00:00" } }, { "txn_id": "BB-UW-0103", "assessment_id": "ra_80697a6b738842e2", "occurred_at": "2026-07-31T13:03:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_836f797e3e" }, "output": { "decision": "approve", "score": 0.091, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.782609+00:00", "fingerprint": "25e6c0c032185ab88e0b5bf1" }, "review": null, "ledger": { "seq": 103, "hash": "7cd1722422bd37b40d84f5a7a86e4e1c2a59c46cc06e71f07304e463586f5d0c", "prev_hash": "e0db991004f132198c09494699efa92ed48c30c384932678a2473b038cd28657", "recorded_at": "2026-07-31T13:04:05+00:00" } }, { "txn_id": "BB-UW-0042", "assessment_id": "ra_f85712b312f14882", "occurred_at": "2026-07-31T16:32:17+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_82f714c462" }, "output": { "decision": "approve", "score": 0.269, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-07-31, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-07-31 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-07-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.627856+00:00", "fingerprint": "807cdfbf36584bfa2bb9804a" }, "review": null, "ledger": { "seq": 42, "hash": "340c6e85b37b172d7be111089541d0213630f636aaead54c5ac0829b0bdbda92", "prev_hash": "9fe4e7f3b3b40ef24b0447863e5df4a7b45a274a2510111be15a251d1d945c3f", "recorded_at": "2026-07-31T16:35:31+00:00" } }, { "txn_id": "BB-UW-0096", "assessment_id": "ra_e426593c7257424a", "occurred_at": "2026-08-02T15:38:44+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9c1a06703f" }, "output": { "decision": "refer", "score": 0.524, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-02, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.762457+00:00", "fingerprint": "39df2df46ec7058f78f418e5" }, "review": null, "ledger": { "seq": 96, "hash": "0f21d921b7505887978174662de6d879b7b325913b077ac913984475dd2a6f9f", "prev_hash": "51f898f3b94a975c010bcbffd1a2effca58c23b1b1524098e78f524528e8f055", "recorded_at": "2026-08-02T15:42:01+00:00" } }, { "txn_id": "BB-UW-0092", "assessment_id": "ra_f66af2e76fa64c74", "occurred_at": "2026-08-04T01:59:09+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_75b6e57418" }, "output": { "decision": "approve", "score": 0.427, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-04, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.753919+00:00", "fingerprint": "9e5bda2fb44cadd001e4a3bd" }, "review": null, "ledger": { "seq": 92, "hash": "ff637bbd3491ccaafa060fb937eaeff5c811ef97d05368f6a7820d2e4735324f", "prev_hash": "6bbecd1e12300ff8a89decfa92af2b1d4b823cb62874fed65bbff494003507bc", "recorded_at": "2026-08-04T02:01:29+00:00" } }, { "txn_id": "BB-UW-0140", "assessment_id": "ra_406f15fd1f844e1e", "occurred_at": "2026-08-04T03:20:07+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_b8387a0a05" }, "output": { "decision": "approve", "score": 0.237, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-04, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-04 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.878417+00:00", "fingerprint": "72333d9f85bea56fb561044b" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 61.0, "reviewer_role": "underwriter", "resolved_at": "2026-08-06T16:26:23.485047+00:00" }, "ledger": { "seq": 140, "hash": "ac3cd1832b2d173f748f7e8e4edb2c2c3f229f59dda30a9aeeac2784b11305e0", "prev_hash": "7297431612e17506ccf327d8de9c6b3a5da6fb067d8026aa5d3c0598bcaa1542", "recorded_at": "2026-08-04T03:23:41+00:00" } }, { "txn_id": "BB-UW-0066", "assessment_id": "ra_601b3a684b174c55", "occurred_at": "2026-08-04T18:08:12+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_4423a1dea2" }, "output": { "decision": "approve", "score": 0.117, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-04, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-08-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.677310+00:00", "fingerprint": "6c694ef7f311910b62c43e85" }, "review": null, "ledger": { "seq": 66, "hash": "f77671b24ff957a3de06f36bef886d5faba0f72eaf135df104216c3f5df18cd6", "prev_hash": "4bc10a5aa51dd7fb3d8a0d10cfb20bf046331a198866824f8ee16966329ea5a9", "recorded_at": "2026-08-04T18:11:58+00:00" } }, { "txn_id": "BB-UW-0022", "assessment_id": "ra_28637fc31281404c", "occurred_at": "2026-08-06T08:24:34+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2c5814b53f" }, "output": { "decision": "approve", "score": 0.216, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.578547+00:00", "fingerprint": "da65fe902b70781cd0ce3e57" }, "review": null, "ledger": { "seq": 22, "hash": "17ad1ed4b2b876f89e1475030cc9cc4ebf6a8ae31affff609f37edfacb0224b6", "prev_hash": "cf76eafe0f1f806378b08932609a7546ccce875bfeb19188c7cb9277bd19150e", "recorded_at": "2026-08-06T08:26:33+00:00" } }, { "txn_id": "BB-UW-0039", "assessment_id": "ra_4f330348cfbc4c1e", "occurred_at": "2026-08-06T11:41:17+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a02885f868" }, "output": { "decision": "approve", "score": 0.192, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.618584+00:00", "fingerprint": "da65fe902b70781cd0ce3e57" }, "review": null, "ledger": { "seq": 39, "hash": "99935d8b892761762b7e6c539631abf1c29b506166abf72fe4df3072200ca525", "prev_hash": "b3eecebb9ae167d01eb9f21951082c0ce0c24b357549c355f2ed9fc44ba7ee02", "recorded_at": "2026-08-06T11:42:50+00:00" } }, { "txn_id": "BB-UW-0072", "assessment_id": "ra_c06c7ea1227b4c46", "occurred_at": "2026-08-06T18:11:54+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_f2e6445bb0" }, "output": { "decision": "approve", "score": 0.108, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.692707+00:00", "fingerprint": "dad018a0b23c13b882c8e105" }, "review": null, "ledger": { "seq": 72, "hash": "c31cac566b1030178f8f9c354ac78aa3b17637c47dbd13b44e47b2783a65f23e", "prev_hash": "d2b67932c53b12959104af96d9278ff6337cf27d5d94b87c856efa9054124276", "recorded_at": "2026-08-06T18:13:18+00:00" } }, { "txn_id": "BB-UW-0043", "assessment_id": "ra_9338e8970821464a", "occurred_at": "2026-08-06T20:25:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e62986a329" }, "output": { "decision": "approve", "score": 0.439, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-06, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.629851+00:00", "fingerprint": "a012f4fc07c103e9a7563185" }, "review": null, "ledger": { "seq": 43, "hash": "1e8d31db99b3bd7f1dee53493c296370d1f366adc5200bbf640779eb4ff21b3b", "prev_hash": "340c6e85b37b172d7be111089541d0213630f636aaead54c5ac0829b0bdbda92", "recorded_at": "2026-08-06T20:27:45+00:00" } }, { "txn_id": "BB-UW-0147", "assessment_id": "ra_7a4f622b3e864ba4", "occurred_at": "2026-08-08T14:14:51+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_cdc08bfcd6" }, "output": { "decision": "refer", "score": 0.46, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-08, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.893879+00:00", "fingerprint": "eca9d51995b433036be3c0ce" }, "review": null, "ledger": { "seq": 147, "hash": "f6e2e9c2d60ca6100b834beb3b2ae482656b5c183d486d83e50641b72e6342c2", "prev_hash": "62574275b17d245129eb50350d11ab61d10aca1776cd964141fe10138a86622e", "recorded_at": "2026-08-08T14:17:28+00:00" } }, { "txn_id": "BB-UW-0071", "assessment_id": "ra_eaaaef70cd91448a", "occurred_at": "2026-08-10T23:34:50+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_32bd328e8b" }, "output": { "decision": "approve", "score": 0.299, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.688398+00:00", "fingerprint": "0924980c4da28fbcc2c8fa46" }, "review": null, "ledger": { "seq": 71, "hash": "d2b67932c53b12959104af96d9278ff6337cf27d5d94b87c856efa9054124276", "prev_hash": "52dccfe65eedb37f144748b78e6f9729c7119a39b5b37e4b153417a24cf443d1", "recorded_at": "2026-08-10T23:38:16+00:00" } }, { "txn_id": "BB-UW-0082", "assessment_id": "ra_d02a1a4613bc4adc", "occurred_at": "2026-08-12T02:22:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_139629078f" }, "output": { "decision": "decline", "score": 0.924, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-13T02:22:13+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-12, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-13 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-12 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.718105+00:00", "fingerprint": "73acf3bdd3a71e0e946e0079" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 14.2, "reviewer_role": "underwriter", "resolved_at": "2026-08-12T16:35:53.322308+00:00" }, "ledger": { "seq": 82, "hash": "2d064db703250e1d9fc5990bd765b7635d39f191d8ffab01f0df91fcc43c5246", "prev_hash": "46545d627d59dc90ddd6fa2e9daed5f004d2aceea39b6ba5aeea6449a941793e", "recorded_at": "2026-08-12T02:25:18+00:00" } }, { "txn_id": "BB-UW-0041", "assessment_id": "ra_64856943dce243bc", "occurred_at": "2026-08-12T03:51:52+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_d9d98d4098" }, "output": { "decision": "approve", "score": 0.207, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-12, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.626136+00:00", "fingerprint": "aa481cc8b0fc3d5c8895f39b" }, "review": null, "ledger": { "seq": 41, "hash": "9fe4e7f3b3b40ef24b0447863e5df4a7b45a274a2510111be15a251d1d945c3f", "prev_hash": "a336e1644abcf3d96613124bb4ead1587d958291082af6599ae774200473cd5f", "recorded_at": "2026-08-12T03:54:11+00:00" } }, { "txn_id": "BB-UW-0055", "assessment_id": "ra_144cbd526a434fc0", "occurred_at": "2026-08-12T16:40:25+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_1ea2e29977" }, "output": { "decision": "approve", "score": 0.362, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-12, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-12 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.653238+00:00", "fingerprint": "2996c00a7bbac18951d6b999" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 11.9, "reviewer_role": "underwriter", "resolved_at": "2026-08-13T04:37:10.474103+00:00" }, "ledger": { "seq": 55, "hash": "9582c471d169be951d9a28434be85ca957e295a52e3884af86f03b014b0ac59e", "prev_hash": "ab3a80cfa4885abdf192fd4ee63efa30a80f0231668351cfd9967821b5778514", "recorded_at": "2026-08-12T16:41:03+00:00" } }, { "txn_id": "BB-UW-0074", "assessment_id": "ra_48b6a89a67b6453d", "occurred_at": "2026-08-13T05:21:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5bf83c61f0" }, "output": { "decision": "approve", "score": 0.422, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.696285+00:00", "fingerprint": "e68fdb7f2d7a542b29fc6695" }, "review": null, "ledger": { "seq": 74, "hash": "f53de172bd571eccdbb2354d5aa30be25f7305da54dead054baad58898e22133", "prev_hash": "4a575f8aed435a86c4a8fbf73aed6b791511e948ac68513798f12bf9258cdbda", "recorded_at": "2026-08-13T05:24:27+00:00" } }, { "txn_id": "BB-UW-0015", "assessment_id": "ra_cd59bd50a0a64d09", "occurred_at": "2026-08-13T06:55:05+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_96791572d5" }, "output": { "decision": "approve", "score": 0.166, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.559206+00:00", "fingerprint": "5ab9d563fb6cb79b36e928d3" }, "review": null, "ledger": { "seq": 15, "hash": "ef47edfe70255c3bbab2fefe6108428405a46af427aaac11fe3b9e4cc4efa9d8", "prev_hash": "30f5bcbfc6339b1302a0237715e11766ed6c393283193471f0856487d2249241", "recorded_at": "2026-08-13T06:58:44+00:00" } }, { "txn_id": "BB-UW-0097", "assessment_id": "ra_39b5bd73a7204ba0", "occurred_at": "2026-08-13T15:52:15+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_b1de1cfb0a" }, "output": { "decision": "approve", "score": 0.204, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.763853+00:00", "fingerprint": "3ac1ce438e72b31ea18321be" }, "review": null, "ledger": { "seq": 97, "hash": "7ef37006fe2e18bb3d431351d05f7675ae29bc9ea9adf62201fb618b8d4772c4", "prev_hash": "0f21d921b7505887978174662de6d879b7b325913b077ac913984475dd2a6f9f", "recorded_at": "2026-08-13T15:55:14+00:00" } }, { "txn_id": "BB-UW-0143", "assessment_id": "ra_fe666f1f2d024f25", "occurred_at": "2026-08-13T20:38:18+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9d43e80903" }, "output": { "decision": "approve", "score": 0.118, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-08-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.885296+00:00", "fingerprint": "3f7e57b6cb1af83b691d4747" }, "review": null, "ledger": { "seq": 143, "hash": "59e5f414b94246103c63d479721d772d206c9706418448b48ce7b615b6d4065b", "prev_hash": "0bbda96f8baca277ecbb1e915e816cdd02c80c0fd80e7b6c501b8673601ce30f", "recorded_at": "2026-08-13T20:41:50+00:00" } }, { "txn_id": "BB-UW-0087", "assessment_id": "ra_543af6e1c86a473d", "occurred_at": "2026-08-14T07:56:02+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_64896913c7" }, "output": { "decision": "approve", "score": 0.204, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.736401+00:00", "fingerprint": "cbd6d79e588d51b8cdcdeb8e" }, "review": null, "ledger": { "seq": 87, "hash": "b2d8f39f914dc0275aa1cb4073b35f7012f8fcc11e56bc811cbe8637a2eab014", "prev_hash": "79c778bbe13efc31b9d0c5eda792c1acb3216efd04e0f76c717d27adf3f4d4c1", "recorded_at": "2026-08-14T07:58:47+00:00" } }, { "txn_id": "BB-UW-0088", "assessment_id": "ra_cf5dbf4746344ae6", "occurred_at": "2026-08-15T23:58:04+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_8f2a228326" }, "output": { "decision": "approve", "score": 0.443, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-15, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.742196+00:00", "fingerprint": "2ae74ab504b2b6691b42c4bc" }, "review": null, "ledger": { "seq": 88, "hash": "eeb0502df7c4baf4c662d87f5a1a90acf8c0c20d2ad7a6a15982f33d56cfdf2f", "prev_hash": "b2d8f39f914dc0275aa1cb4073b35f7012f8fcc11e56bc811cbe8637a2eab014", "recorded_at": "2026-08-15T23:58:53+00:00" } }, { "txn_id": "BB-UW-0139", "assessment_id": "ra_356b50367f594dda", "occurred_at": "2026-08-16T08:46:55+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e3bb464d24" }, "output": { "decision": "approve", "score": 0.086, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-16, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-08-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.876039+00:00", "fingerprint": "d7de53feff9ca52ebbe05cdf" }, "review": null, "ledger": { "seq": 139, "hash": "7297431612e17506ccf327d8de9c6b3a5da6fb067d8026aa5d3c0598bcaa1542", "prev_hash": "e1ef76c3b8be36a57044b9d040db09abae8087ddcd61d0ad661ba8f508ee905e", "recorded_at": "2026-08-16T08:49:23+00:00" } }, { "txn_id": "BB-UW-0003", "assessment_id": "ra_af5b43b466f9422d", "occurred_at": "2026-08-18T20:08:09+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_fa59001ac9" }, "output": { "decision": "approve", "score": 0.348, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-18, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-18 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.531796+00:00", "fingerprint": "da616fed4f00cc3dd673bb6f" }, "review": null, "ledger": { "seq": 3, "hash": "38607c1d1af3bcf9d3628ad652b3c3aa5f384d9a6580ce45d18b3a941a82c6da", "prev_hash": "bf4d3dc9cbd835c9279703284fd3729c67ecadca50e8cec8815980c9497618d7", "recorded_at": "2026-08-18T20:09:28+00:00" } }, { "txn_id": "BB-UW-0111", "assessment_id": "ra_e7aba153a5f14f67", "occurred_at": "2026-08-18T21:12:43+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7586ffc468" }, "output": { "decision": "decline", "score": 0.758, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-22T21:12:43+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-18, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-22 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-18 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.815133+00:00", "fingerprint": "cf2828f81e2cc949a3589dd0" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 111, "hash": "9580e44fba902edc18378abc355beaf8168625777dff1814948d90446b26497e", "prev_hash": "4c8c3a648814b180765c3b35c9f788df11a53afc3fc079d2477fb2bcdbb374b2", "recorded_at": "2026-08-18T21:13:20+00:00" } }, { "txn_id": "BB-UW-0133", "assessment_id": "ra_8881d7fe677e4322", "occurred_at": "2026-08-19T07:45:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a28e09d843" }, "output": { "decision": "approve", "score": 0.355, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.863866+00:00", "fingerprint": "a5bf09fe62745fc0d316a575" }, "review": null, "ledger": { "seq": 133, "hash": "1e68cbaebc8bff570e82fe39ecd5e594c88d88c32a895b47aca1a576f2df8f5d", "prev_hash": "d5bad2ac3373823956d2dd21ad68382aeb15e7ca898cd92dbeb8e01b51790d0b", "recorded_at": "2026-08-19T07:48:59+00:00" } }, { "txn_id": "BB-UW-0028", "assessment_id": "ra_845e2787f8c243ee", "occurred_at": "2026-08-19T21:54:09+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_f76a1e860d" }, "output": { "decision": "approve", "score": 0.099, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.593523+00:00", "fingerprint": "d502ebf5130f3a118a3225e4" }, "review": null, "ledger": { "seq": 28, "hash": "6f064ddeab61e8727ce83cc22fe51892e8954e2d2074131994bd03dc97941227", "prev_hash": "5afe4da57af98212fe9ba24eae9b1521be5296853307ae2dd766c570347b1eed", "recorded_at": "2026-08-19T21:56:14+00:00" } }, { "txn_id": "BB-UW-0069", "assessment_id": "ra_98325c8515e64bc7", "occurred_at": "2026-08-19T22:46:04+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_c1542a6174" }, "output": { "decision": "decline", "score": 0.725, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-21T22:46:04+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-19, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-08-21 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.682851+00:00", "fingerprint": "60487dd0f114104796bfd2e6" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 21.8, "reviewer_role": "underwriter", "resolved_at": "2026-08-20T20:36:47.899145+00:00" }, "ledger": { "seq": 69, "hash": "ab0eda9df0814e3a92d0119d45255f30107bcb6346132e45a6ffc2183a0447b0", "prev_hash": "5962da7082eb39ffd6c3d26940b77858f21931c8f38c52e7729535a733ce40e4", "recorded_at": "2026-08-19T22:47:42+00:00" } }, { "txn_id": "BB-UW-0062", "assessment_id": "ra_c28fa0650915484b", "occurred_at": "2026-08-19T23:59:27+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_f2f6c5bbc1" }, "output": { "decision": "decline", "score": 0.877, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 6 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-08-19, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-08-19 and recorded 6 finding(s); the most serious is 'Adverse action based on a consumer report with no FCRA notice recorded' (critical, L0).", "findings": [ { "rule_id": "us.fcra.adverse-action-notice", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "critical", "title": "Adverse action based on a consumer report with no FCRA notice recorded", "required_action": "Hold the outcome for compliance review; generate the notice with every required element before the decision is communicated, and record notice_sent_at on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "us-fcra-1681m-a", "label": "15 U.S.C. \u00a7 1681m(a) (FCRA \u00a7 615(a)) \u2014 duties of users taking adverse actions on the basis of a consumer report", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681m", "verified": true }, { "provision_id": "us-fcra-1681a-k-1-b-i", "label": "15 U.S.C. \u00a7 1681a(k)(1)(B)(i) \u2014 'adverse action' in connection with insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681a", "verified": true } ] }, { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "layer": "L5", "judgment_basis": "precedent", "advisory": true, "severity": "medium", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "required_action": "Confirm the notice and document deadlines before the cancellation or decline takes effect; record the notice on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "precedent-pa-geico-2026", "label": "Pennsylvania Attorney General agreement with GEICO (announced May 2026) \u2014 AI-enabled underwriting review and cancellation without adequate notice", "source_url": "https://www.attorneygeneral.gov/?p=33518", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.668650+00:00", "fingerprint": "a91de27517177fd72b6c57d1" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 14.1, "reviewer_role": "underwriter", "resolved_at": "2026-08-20T14:07:24.501697+00:00" }, "ledger": { "seq": 62, "hash": "1550deee986d16c9223ae32719aba3411be4686f78de513475c6c4e9ba7a5075", "prev_hash": "f86b262c1ef79c94a8b45f55a2314c8e9d60e31dfe9299412f620cfc36e29332", "recorded_at": "2026-08-20T00:02:30+00:00" } }, { "txn_id": "BB-UW-0083", "assessment_id": "ra_5ca6e601edae433a", "occurred_at": "2026-08-21T17:27:48+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5fe983f1ec" }, "output": { "decision": "decline", "score": 0.761, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-26T17:27:48+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-21, uw-triage v2.4 (an internally developed system) produced 'decline' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-08-26 with specific reasons. RegBench applied 10 rule(s) in force on 2026-08-21 and recorded 2 finding(s); the most serious is 'Adverse outcome from a high-risk AI system with no human review' (medium, L6).", "findings": [ { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.719887+00:00", "fingerprint": "abd6d451be6a336adae08704" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 62.3, "reviewer_role": "underwriter", "resolved_at": "2026-08-24T07:48:23.436616+00:00" }, "ledger": { "seq": 83, "hash": "d6fc1cc83e96f74df56b388266f57107b586c5ee87dd0add79f0bf4b78bf17dd", "prev_hash": "2d064db703250e1d9fc5990bd765b7635d39f191d8ffab01f0df91fcc43c5246", "recorded_at": "2026-08-21T17:29:08+00:00" } }, { "txn_id": "BB-UW-0081", "assessment_id": "ra_a581185a252e43e7", "occurred_at": "2026-08-21T21:44:40+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6ec459318e" }, "output": { "decision": "approve", "score": 0.398, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-21, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-21 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.716566+00:00", "fingerprint": "f9215cd0599a4015828c1d39" }, "review": null, "ledger": { "seq": 81, "hash": "46545d627d59dc90ddd6fa2e9daed5f004d2aceea39b6ba5aeea6449a941793e", "prev_hash": "b333ef195377342083269ad98c0463758d0e6586a76ecf2416fd52b41d97c766", "recorded_at": "2026-08-21T21:47:54+00:00" } }, { "txn_id": "BB-UW-0128", "assessment_id": "ra_ea76c3cb4c7f4664", "occurred_at": "2026-08-22T02:06:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_64744be5da" }, "output": { "decision": "approve", "score": 0.133, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-22 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.852384+00:00", "fingerprint": "9ee2a18efecae25df7f0e7ce" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 18.1, "reviewer_role": "underwriter", "resolved_at": "2026-08-22T20:18:24.176328+00:00" }, "ledger": { "seq": 128, "hash": "63c2a0bbabfb55f92050df6655cf38f72006a707007179d2ccafadaead7b1a83", "prev_hash": "9fe4d23cc6d8c5e978ef2d1bb2a481b149a087b578fa2fabeeb1c2c2f1ddebca", "recorded_at": "2026-08-22T02:10:01+00:00" } }, { "txn_id": "BB-UW-0095", "assessment_id": "ra_ce0ed2a967704c96", "occurred_at": "2026-08-23T03:35:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_c47053ce0b" }, "output": { "decision": "approve", "score": 0.228, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-23 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.760867+00:00", "fingerprint": "8b42a16f7964dfeece027c12" }, "review": { "status": "resolved", "outcome": "escalated", "hours_to_resolve": 27.9, "reviewer_role": "underwriter", "resolved_at": "2026-08-24T07:30:35.130554+00:00" }, "ledger": { "seq": 95, "hash": "51f898f3b94a975c010bcbffd1a2effca58c23b1b1524098e78f524528e8f055", "prev_hash": "e672ed609145028d64e2748dca06a58acfc9e3fcdbb217e865bb684f320c1789", "recorded_at": "2026-08-23T03:38:45+00:00" } }, { "txn_id": "BB-UW-0038", "assessment_id": "ra_b546f96f6d7b457c", "occurred_at": "2026-08-23T13:08:37+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_dc57fbeeeb" }, "output": { "decision": "approve", "score": 0.27, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-08-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.617032+00:00", "fingerprint": "2d293f8607852758e764b3b4" }, "review": null, "ledger": { "seq": 38, "hash": "b3eecebb9ae167d01eb9f21951082c0ce0c24b357549c355f2ed9fc44ba7ee02", "prev_hash": "80fcb2a9e7b3d862345ce98ffea91f5f9138235672a783c1aea474f345fcdc3e", "recorded_at": "2026-08-23T13:09:08+00:00" } }, { "txn_id": "BB-UW-0141", "assessment_id": "ra_80238885e0044364", "occurred_at": "2026-08-26T08:17:39+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_17b1ae1359" }, "output": { "decision": "decline", "score": 0.83, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-08-30T08:17:39+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-26, uw-triage v2.4 (an internally developed system) produced 'decline' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-08-30 with reasons stated. RegBench applied 6 rule(s) in force on 2026-08-26 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.881455+00:00", "fingerprint": "e1f1a795f280e08ca850437a" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 19.8, "reviewer_role": "underwriter", "resolved_at": "2026-08-27T04:09:53.532768+00:00" }, "ledger": { "seq": 141, "hash": "5dfd89edecd1e6838c827c6e3954e754e8ffc834aac2f49a6d293ae3bb795ae7", "prev_hash": "ac3cd1832b2d173f748f7e8e4edb2c2c3f229f59dda30a9aeeac2784b11305e0", "recorded_at": "2026-08-26T08:20:49+00:00" } }, { "txn_id": "BB-UW-0033", "assessment_id": "ra_394ce67757ed4380", "occurred_at": "2026-08-28T16:32:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6c075f0e1d" }, "output": { "decision": "approve", "score": 0.146, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-28, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-28 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-28", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.602258+00:00", "fingerprint": "0df45f25a37091f54c260f56" }, "review": null, "ledger": { "seq": 33, "hash": "e34e2d531f4aa0b9c5f02d614dd38bc7e2580f89defb602ca635a307f92beb3e", "prev_hash": "1ee83939bbd9d6a2fec49ae919a895fba8b03a90407d6c92aaccd0905a6fee2f", "recorded_at": "2026-08-28T16:36:01+00:00" } }, { "txn_id": "BB-UW-0148", "assessment_id": "ra_1234f2fc26bd4d7d", "occurred_at": "2026-08-30T13:00:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_fc867735b0" }, "output": { "decision": "approve", "score": 0.358, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-08-30, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-08-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-08-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.895742+00:00", "fingerprint": "50d8390c5172d495ba57f38c" }, "review": null, "ledger": { "seq": 148, "hash": "ed5e042b94915834033dc792ce5b1b77be3d5885a8bdc5b36edc1c1e7cad6a2b", "prev_hash": "f6e2e9c2d60ca6100b834beb3b2ae482656b5c183d486d83e50641b72e6342c2", "recorded_at": "2026-08-30T13:04:19+00:00" } }, { "txn_id": "BB-UW-0059", "assessment_id": "ra_aea8a2fd14b64f46", "occurred_at": "2026-08-31T17:08:26+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7a9b4a8d0a" }, "output": { "decision": "decline", "score": 0.71, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-03T17:08:26+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-08-31, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-03 with specific reasons. RegBench applied 11 rule(s) in force on 2026-08-31 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-08-31", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.661954+00:00", "fingerprint": "9fbf118fe48431553a475c4b" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 10.8, "reviewer_role": "underwriter", "resolved_at": "2026-09-01T03:59:46.901219+00:00" }, "ledger": { "seq": 59, "hash": "5c7d48c865a05bc7e5062c046be16df0db205e3b510233db68bb1ba3b1e92211", "prev_hash": "f49144eb18bf330867d0fbaf15fd5040467abdde904041a55e6f832cb839dd93", "recorded_at": "2026-08-31T17:11:30+00:00" } }, { "txn_id": "BB-UW-0064", "assessment_id": "ra_a09abd0fc8914ed0", "occurred_at": "2026-09-02T06:09:15+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_ff5d1e70ff" }, "output": { "decision": "approve", "score": 0.126, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-02, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.674064+00:00", "fingerprint": "607190dfd5caac5e8235b094" }, "review": null, "ledger": { "seq": 64, "hash": "7062a019656d4180156b841993d8b18d7e5af371e59638eb60c110a2ea816f76", "prev_hash": "6306bce53afdd844c4067fc254adf728a2ebe3c2a487798511200972f31b969f", "recorded_at": "2026-09-02T06:10:23+00:00" } }, { "txn_id": "BB-UW-0024", "assessment_id": "ra_3588b97b8ea04da6", "occurred_at": "2026-09-03T03:58:19+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_8b6872d96b" }, "output": { "decision": "approve", "score": 0.434, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-03, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.582407+00:00", "fingerprint": "65be4cd094865d5691a76d48" }, "review": null, "ledger": { "seq": 24, "hash": "1fd83882254587608e93f17509fef55156720474276d280ffa671f86d6630bac", "prev_hash": "3e1f6a418069dab68648a6fd83dfbfcc152dd674142231388d34061cd89740ca", "recorded_at": "2026-09-03T04:00:43+00:00" } }, { "txn_id": "BB-UW-0031", "assessment_id": "ra_b189e83218594559", "occurred_at": "2026-09-03T04:27:56+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_ba4a205268" }, "output": { "decision": "approve", "score": 0.357, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-03, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-03 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-03", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.598801+00:00", "fingerprint": "58739ddbd0bab3b3770ead9a" }, "review": null, "ledger": { "seq": 31, "hash": "f921d3dc6627505c6d1cf012200e30c182fed0cbcdec4670fecb107e7f974e17", "prev_hash": "f971da6c6e953e522bb4dfacbc951797e7dfc60de8e3947b8c512df993db46b9", "recorded_at": "2026-09-03T04:28:55+00:00" } }, { "txn_id": "BB-UW-0019", "assessment_id": "ra_22b9a9b1e4414462", "occurred_at": "2026-09-04T07:49:31+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9975750c2b" }, "output": { "decision": "approve", "score": 0.064, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-04, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.570318+00:00", "fingerprint": "fbe5fcfc14f6e655fcb017df" }, "review": null, "ledger": { "seq": 19, "hash": "a4f4f78a3658f13cc57deb539d01ded356b28c0e4165830fbcfcbb79a73511bd", "prev_hash": "4b15d24efa3264740de91299894e5cd74fb9ee5826f05d1f1c36fab992576cf4", "recorded_at": "2026-09-04T07:52:34+00:00" } }, { "txn_id": "BB-UW-0113", "assessment_id": "ra_6813c77af7cb42d3", "occurred_at": "2026-09-04T07:55:20+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a567eff5ed" }, "output": { "decision": "approve", "score": 0.093, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-04, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-04 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-04", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.821003+00:00", "fingerprint": "f6f749bc2500d445284166b9" }, "review": null, "ledger": { "seq": 113, "hash": "7fa43cd548fec655412206c628fbc6208d71155603b015964c1f6f9b650a327f", "prev_hash": "798964458d908cf91683265745a15e3dbd3f683901ee7060ce3c5dbd1ccf3c6a", "recorded_at": "2026-09-04T07:56:24+00:00" } }, { "txn_id": "BB-UW-0032", "assessment_id": "ra_ab5a44d6aa44475a", "occurred_at": "2026-09-05T12:05:25+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_8f328bf8ae" }, "output": { "decision": "approve", "score": 0.099, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-05, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-05 and recorded 1 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-05", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.600682+00:00", "fingerprint": "b99a60b608d5919b5c288290" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 69.0, "reviewer_role": "underwriter", "resolved_at": "2026-09-08T09:08:22.874094+00:00" }, "ledger": { "seq": 32, "hash": "1ee83939bbd9d6a2fec49ae919a895fba8b03a90407d6c92aaccd0905a6fee2f", "prev_hash": "f921d3dc6627505c6d1cf012200e30c182fed0cbcdec4670fecb107e7f974e17", "recorded_at": "2026-09-05T12:06:11+00:00" } }, { "txn_id": "BB-UW-0029", "assessment_id": "ra_a99cc3384d414306", "occurred_at": "2026-09-07T00:32:53+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_c750a22a1b" }, "output": { "decision": "approve", "score": 0.065, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-07, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.595299+00:00", "fingerprint": "29a874679d449e80d8fabc63" }, "review": null, "ledger": { "seq": 29, "hash": "0dec72ffad90217c962eba397fbf68a11a54ede4c23f95e998bc98cdef595ac3", "prev_hash": "6f064ddeab61e8727ce83cc22fe51892e8954e2d2074131994bd03dc97941227", "recorded_at": "2026-09-07T00:35:44+00:00" } }, { "txn_id": "BB-UW-0135", "assessment_id": "ra_7269f5c718e94275", "occurred_at": "2026-09-08T02:45:52+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_1ff51ad83a" }, "output": { "decision": "approve", "score": 0.133, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-08, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.867259+00:00", "fingerprint": "ab1dfc3a4c1c0d95df340f18" }, "review": null, "ledger": { "seq": 135, "hash": "cfa7b35b29970665018de771f5f8fb93895ad3f4f45e61bf7a10acdef27becab", "prev_hash": "2c9d311c3cae1d74acd2e79341fc611b61a2ac90b5c735b7ff6893b27598ffff", "recorded_at": "2026-09-08T02:47:32+00:00" } }, { "txn_id": "BB-UW-0115", "assessment_id": "ra_cb958d7c762c4776", "occurred_at": "2026-09-08T22:30:27+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7769d1aab6" }, "output": { "decision": "approve", "score": 0.266, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-08, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.827735+00:00", "fingerprint": "c58bef7d8bed43abbae34c72" }, "review": null, "ledger": { "seq": 115, "hash": "16cdfde96b972988e1074bb1e051c4162b16db8bf0194320b26cc42d1982e799", "prev_hash": "ff9d84c9fcd9bcac0e491d28572a73a4f5b5f7cc0b07e7cbfc9ad06b747b3e81", "recorded_at": "2026-09-08T22:31:49+00:00" } }, { "txn_id": "BB-UW-0073", "assessment_id": "ra_8245cfd02e314277", "occurred_at": "2026-09-09T22:04:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_0265f71bd5" }, "output": { "decision": "approve", "score": 0.192, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.694839+00:00", "fingerprint": "eca2d193009812bc89245ef4" }, "review": null, "ledger": { "seq": 73, "hash": "4a575f8aed435a86c4a8fbf73aed6b791511e948ac68513798f12bf9258cdbda", "prev_hash": "c31cac566b1030178f8f9c354ac78aa3b17637c47dbd13b44e47b2783a65f23e", "recorded_at": "2026-09-09T22:08:40+00:00" } }, { "txn_id": "BB-UW-0117", "assessment_id": "ra_c2044be24c7648e3", "occurred_at": "2026-09-10T06:54:18+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_deef659464" }, "output": { "decision": "approve", "score": 0.295, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.830861+00:00", "fingerprint": "50c1ed8718fe24eec1174f10" }, "review": null, "ledger": { "seq": 117, "hash": "0095f5b056445b0f9565b39b0b8a2b7c9901ca326822b70983ddc45669d06955", "prev_hash": "390ee65e2953424d88e867a0062b0cff55e38611f08f8c476a302dbb5daddf4d", "recorded_at": "2026-09-10T06:56:27+00:00" } }, { "txn_id": "BB-UW-0138", "assessment_id": "ra_c12376f217e14185", "occurred_at": "2026-09-10T11:03:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_37dfc81c69" }, "output": { "decision": "approve", "score": 0.332, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.873742+00:00", "fingerprint": "50c1ed8718fe24eec1174f10" }, "review": null, "ledger": { "seq": 138, "hash": "e1ef76c3b8be36a57044b9d040db09abae8087ddcd61d0ad661ba8f508ee905e", "prev_hash": "f703a643c649bbf1a9e771f1d8aaee2f496367febd7c1fea3ee53df38fa5adbe", "recorded_at": "2026-09-10T11:07:34+00:00" } }, { "txn_id": "BB-UW-0056", "assessment_id": "ra_a4baee8a4de040b2", "occurred_at": "2026-09-10T17:41:19+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_8e14388e02" }, "output": { "decision": "approve", "score": 0.088, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.656149+00:00", "fingerprint": "e64d5457463489a3bfe1563e" }, "review": null, "ledger": { "seq": 56, "hash": "09bcbd50ffebfa12ef23a4ec7447bee628ae4f2620045ea1b60b5f8fa2a96fa0", "prev_hash": "9582c471d169be951d9a28434be85ca957e295a52e3884af86f03b014b0ac59e", "recorded_at": "2026-09-10T17:44:45+00:00" } }, { "txn_id": "BB-UW-0012", "assessment_id": "ra_fe4d7abaa2a345c8", "occurred_at": "2026-09-10T20:55:30+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_3118bee074" }, "output": { "decision": "approve", "score": 0.056, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-10, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-10 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-10", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.550132+00:00", "fingerprint": "d7e8c74af62de4d0ec4cced4" }, "review": null, "ledger": { "seq": 12, "hash": "ce337d78c19c99fe66216ee1999520fc41b7be9a999b03efaf3f77fd9a36857a", "prev_hash": "37fb57ef1172d4dc84c7f5fc9b9114a42622f23fcaa10d2fb807a9af689b31fe", "recorded_at": "2026-09-10T20:56:51+00:00" } }, { "txn_id": "BB-UW-0094", "assessment_id": "ra_1abf95d6516d40d5", "occurred_at": "2026-09-13T00:14:05+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a0a22e9caa" }, "output": { "decision": "approve", "score": 0.383, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.759358+00:00", "fingerprint": "115e137adfab7ed9f39304ba" }, "review": null, "ledger": { "seq": 94, "hash": "e672ed609145028d64e2748dca06a58acfc9e3fcdbb217e865bb684f320c1789", "prev_hash": "6105350864a44d85b3dc0fc5e5e4992ac706ac3866b89fd96fd76b0cf84574ed", "recorded_at": "2026-09-13T00:15:56+00:00" } }, { "txn_id": "BB-UW-0102", "assessment_id": "ra_11429f9f1a8a48f0", "occurred_at": "2026-09-14T05:27:51+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e49f70b8ba" }, "output": { "decision": "decline", "score": 0.857, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-15T05:27:51+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-14, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-15 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-14 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.778684+00:00", "fingerprint": "3133cdd37d11c590eafef186" }, "review": { "status": "open", "outcome": null, "hours_to_resolve": null, "reviewer_role": null, "resolved_at": null }, "ledger": { "seq": 102, "hash": "e0db991004f132198c09494699efa92ed48c30c384932678a2473b038cd28657", "prev_hash": "19ad4e4907b466dfa89302f042f1495f6d24cd8651977cc31d806420c3afbc41", "recorded_at": "2026-09-14T05:30:45+00:00" } }, { "txn_id": "BB-UW-0119", "assessment_id": "ra_5b6e967f52c44f0d", "occurred_at": "2026-09-14T08:53:24+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_23096b4ef8" }, "output": { "decision": "approve", "score": 0.353, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-09-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.835186+00:00", "fingerprint": "ec0124762d6ed6eac10c21d8" }, "review": null, "ledger": { "seq": 119, "hash": "e2c4ad275b148e32e8b3062da6cab3245d958cf131b25a76e71e3f185f75854f", "prev_hash": "c11caa235cff8ade9f6a735859d9d81b168d4ef7b721de4e5b1a2d5440c6d650", "recorded_at": "2026-09-14T08:55:35+00:00" } }, { "txn_id": "BB-UW-0060", "assessment_id": "ra_24d869accc5445a3", "occurred_at": "2026-09-16T15:47:24+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_374dd75f67" }, "output": { "decision": "decline", "score": 0.903, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-21T15:47:24+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-16, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-21 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-16 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.663970+00:00", "fingerprint": "7471afcd7eb414575760b253" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 24.9, "reviewer_role": "underwriter", "resolved_at": "2026-09-17T16:43:08.725400+00:00" }, "ledger": { "seq": 60, "hash": "0740718325691e0ca44b6e793f743da597f6bc0bb1f576aafd997261f71b1174", "prev_hash": "5c7d48c865a05bc7e5062c046be16df0db205e3b510233db68bb1ba3b1e92211", "recorded_at": "2026-09-16T15:50:48+00:00" } }, { "txn_id": "BB-UW-0052", "assessment_id": "ra_4f38b529191a4167", "occurred_at": "2026-09-16T21:56:11+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_0566426014" }, "output": { "decision": "approve", "score": 0.088, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-16, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.648217+00:00", "fingerprint": "32629e6723797558ad746bdd" }, "review": null, "ledger": { "seq": 52, "hash": "b87b3d77cf9a11f79cb9d7b5796ac8db0d7f80d85af01922f8cb3da954f73106", "prev_hash": "cc9a0291b0ba6381755cf377815048221200676655b694cf289b51bcbce42751", "recorded_at": "2026-09-16T21:58:59+00:00" } }, { "txn_id": "BB-UW-0068", "assessment_id": "ra_9b132fdcf3974d74", "occurred_at": "2026-09-17T23:48:23+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e627af415f" }, "output": { "decision": "approve", "score": 0.053, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-17, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-17 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-17", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.680613+00:00", "fingerprint": "dd6b0d79ffc299b1204c0e26" }, "review": null, "ledger": { "seq": 68, "hash": "5962da7082eb39ffd6c3d26940b77858f21931c8f38c52e7729535a733ce40e4", "prev_hash": "a97d2743f249fceba6f4c3cba048317dbbeab298f650c964c9856f41a683f78c", "recorded_at": "2026-09-17T23:50:47+00:00" } }, { "txn_id": "BB-UW-0067", "assessment_id": "ra_7bb2dc916be34476", "occurred_at": "2026-09-19T08:59:38+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_9fa4275452" }, "output": { "decision": "approve", "score": 0.138, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-19, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-19 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-19", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.678708+00:00", "fingerprint": "84e8dab5551ebcce469d6ced" }, "review": null, "ledger": { "seq": 67, "hash": "a97d2743f249fceba6f4c3cba048317dbbeab298f650c964c9856f41a683f78c", "prev_hash": "f77671b24ff957a3de06f36bef886d5faba0f72eaf135df104216c3f5df18cd6", "recorded_at": "2026-09-19T09:01:44+00:00" } }, { "txn_id": "BB-UW-0058", "assessment_id": "ra_38ef0b1ea6a54f99", "occurred_at": "2026-09-20T11:10:40+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2f0b62fb05" }, "output": { "decision": "approve", "score": 0.427, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-20, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.660449+00:00", "fingerprint": "4f59b873be4a9a47af05a3dd" }, "review": null, "ledger": { "seq": 58, "hash": "f49144eb18bf330867d0fbaf15fd5040467abdde904041a55e6f832cb839dd93", "prev_hash": "1e9e6c6424ea106167b9f01871307ba3edbdbf2feea8472ab1c24f442b8f1304", "recorded_at": "2026-09-20T11:12:28+00:00" } }, { "txn_id": "BB-UW-0005", "assessment_id": "ra_606cda69554f4709", "occurred_at": "2026-09-20T23:40:34+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6617371472" }, "output": { "decision": "decline", "score": 0.819, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-24T23:40:34+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-20, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-24 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-20 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.535239+00:00", "fingerprint": "7002a7a2974c68fb8c2e5fba" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 9.0, "reviewer_role": "underwriter", "resolved_at": "2026-09-21T08:43:42.326394+00:00" }, "ledger": { "seq": 5, "hash": "875aa9f6fb59c10bab07bf31c0af277c87bcf32048b882d96446ce157b641b61", "prev_hash": "abc560a63897b1057ff0140eac16b9202e5b370c45472d0f2f33fb74243431cd", "recorded_at": "2026-09-20T23:41:42+00:00" } }, { "txn_id": "BB-UW-0030", "assessment_id": "ra_25cf745f1cab4539", "occurred_at": "2026-09-21T17:52:53+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_adfaa7176f" }, "output": { "decision": "decline", "score": 0.891, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-24T17:52:53+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-21, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-09-24 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-21 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-21", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.597095+00:00", "fingerprint": "230a0bcc0354129468994bb4" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 18.2, "reviewer_role": "underwriter", "resolved_at": "2026-09-22T12:05:33.023911+00:00" }, "ledger": { "seq": 30, "hash": "f971da6c6e953e522bb4dfacbc951797e7dfc60de8e3947b8c512df993db46b9", "prev_hash": "0dec72ffad90217c962eba397fbf68a11a54ede4c23f95e998bc98cdef595ac3", "recorded_at": "2026-09-21T17:55:33+00:00" } }, { "txn_id": "BB-UW-0108", "assessment_id": "ra_73dd50c1d7774759", "occurred_at": "2026-09-22T03:19:20+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_11a3ea5ba2" }, "output": { "decision": "approve", "score": 0.094, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.803609+00:00", "fingerprint": "241082ee6da37b9f7e178429" }, "review": null, "ledger": { "seq": 108, "hash": "9039f67c13b0328e973cc3948d76f2c97c27054d122b78be615d248d69e52888", "prev_hash": "8523bf668e3417abaf0fb0f8d7cd0fe7990747bdf842e9277b70d34b26d522fe", "recorded_at": "2026-09-22T03:22:00+00:00" } }, { "txn_id": "BB-UW-0063", "assessment_id": "ra_a10a7b2bfa114ade", "occurred_at": "2026-09-22T07:13:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_db820e8925" }, "output": { "decision": "approve", "score": 0.158, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.670815+00:00", "fingerprint": "cb5c0e3f4f9c7a1cc74cd378" }, "review": null, "ledger": { "seq": 63, "hash": "6306bce53afdd844c4067fc254adf728a2ebe3c2a487798511200972f31b969f", "prev_hash": "1550deee986d16c9223ae32719aba3411be4686f78de513475c6c4e9ba7a5075", "recorded_at": "2026-09-22T07:17:07+00:00" } }, { "txn_id": "BB-UW-0007", "assessment_id": "ra_099c92467c4a402e", "occurred_at": "2026-09-22T09:11:36+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2478549014" }, "output": { "decision": "decline", "score": 0.875, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-09-27T09:11:36+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-22, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-09-27 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-22 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.541917+00:00", "fingerprint": "f0d8017224cce832d1cdd3d0" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 14.5, "reviewer_role": "underwriter", "resolved_at": "2026-09-22T23:47:12.659488+00:00" }, "ledger": { "seq": 7, "hash": "7ac49466b3155551da453e69d5002d5a7d102b21f36ac555280db56724d66549", "prev_hash": "997247e1d1232831d722095f4efef3b3f92fb8d830baefeee20c0d13025feb8c", "recorded_at": "2026-09-22T09:14:51+00:00" } }, { "txn_id": "BB-UW-0116", "assessment_id": "ra_ed5d57929cc34a2c", "occurred_at": "2026-09-22T20:07:34+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6242f9f17e" }, "output": { "decision": "approve", "score": 0.051, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.829455+00:00", "fingerprint": "5f022ff398d9e7af6b8d9fb2" }, "review": null, "ledger": { "seq": 116, "hash": "390ee65e2953424d88e867a0062b0cff55e38611f08f8c476a302dbb5daddf4d", "prev_hash": "16cdfde96b972988e1074bb1e051c4162b16db8bf0194320b26cc42d1982e799", "recorded_at": "2026-09-22T20:10:20+00:00" } }, { "txn_id": "BB-UW-0134", "assessment_id": "ra_a19c14151bf94bfc", "occurred_at": "2026-09-23T03:43:11+00:00", "state": "CO", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_ee77396758" }, "output": { "decision": "approve", "score": 0.108, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 6 rule(s) in force on 2026-09-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.865686+00:00", "fingerprint": "a1e02bc0dba96e81b2ddac68" }, "review": null, "ledger": { "seq": 134, "hash": "2c9d311c3cae1d74acd2e79341fc611b61a2ac90b5c735b7ff6893b27598ffff", "prev_hash": "1e68cbaebc8bff570e82fe39ecd5e594c88d88c32a895b47aca1a576f2df8f5d", "recorded_at": "2026-09-23T03:46:26+00:00" } }, { "txn_id": "BB-UW-0010", "assessment_id": "ra_bb377b0a159e4e37", "occurred_at": "2026-09-23T07:39:21+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a98ece78b0" }, "output": { "decision": "approve", "score": 0.417, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.546632+00:00", "fingerprint": "7826fc18fe858b3d9495c9cf" }, "review": null, "ledger": { "seq": 10, "hash": "3d723048b894580fd6a2e1ebe9f63e5f45993f484e0ca19b73cecd70c4ed8500", "prev_hash": "6ed3f6a18c2c36c6be48d7c776864f1a5c6ba512e114d6051ccc0a7077f7ead3", "recorded_at": "2026-09-23T07:40:34+00:00" } }, { "txn_id": "BB-UW-0130", "assessment_id": "ra_2a821bf7a5bb45c6", "occurred_at": "2026-09-23T11:17:52+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e0170f4449" }, "output": { "decision": "approve", "score": 0.062, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.857864+00:00", "fingerprint": "de042da57cd2a76681ea96bc" }, "review": null, "ledger": { "seq": 130, "hash": "35f833b7b4f5c05095e0886e5deb6d52e71513055efce85d5eac7128b6964c6f", "prev_hash": "0b94a501a6a47c795540fd4c603f3006d38d5ccfca1446028255c842045c30d8", "recorded_at": "2026-09-23T11:20:00+00:00" } }, { "txn_id": "BB-UW-0120", "assessment_id": "ra_8b1d092a4357487d", "occurred_at": "2026-09-24T16:04:54+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_ee30aef2b6" }, "output": { "decision": "approve", "score": 0.422, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-24, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-24 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-24", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.836437+00:00", "fingerprint": "5b30f94accfbac702c33279c" }, "review": null, "ledger": { "seq": 120, "hash": "9ccc5cb2a91ee75121dbd3d32410f28772f03128ff9743e75b1d9542602b1d37", "prev_hash": "e2c4ad275b148e32e8b3062da6cab3245d958cf131b25a76e71e3f185f75854f", "recorded_at": "2026-09-24T16:07:43+00:00" } }, { "txn_id": "BB-UW-0051", "assessment_id": "ra_9b761c6b184e4cea", "occurred_at": "2026-09-25T07:16:59+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_25437ccafb" }, "output": { "decision": "approve", "score": 0.107, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-25, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.646289+00:00", "fingerprint": "4197e26ba9d458773a78eb3d" }, "review": null, "ledger": { "seq": 51, "hash": "cc9a0291b0ba6381755cf377815048221200676655b694cf289b51bcbce42751", "prev_hash": "b37f14ca62223a385d4a8f06d9069e46cb56db1fcb2f1a7339bd5de461b7b244", "recorded_at": "2026-09-25T07:18:20+00:00" } }, { "txn_id": "BB-UW-0018", "assessment_id": "ra_36d3d43026b04219", "occurred_at": "2026-09-26T09:48:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_e95377899a" }, "output": { "decision": "approve", "score": 0.372, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-26 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.568366+00:00", "fingerprint": "34ffd41f3837eefb3798541b" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 17.5, "reviewer_role": "underwriter", "resolved_at": "2026-09-27T03:22:58.777864+00:00" }, "ledger": { "seq": 18, "hash": "4b15d24efa3264740de91299894e5cd74fb9ee5826f05d1f1c36fab992576cf4", "prev_hash": "3a3dd6b1fbfd7b19ddd0d74b21758584f8116ba76a1146a814c0d8bef3bdf13d", "recorded_at": "2026-09-26T09:50:02+00:00" } }, { "txn_id": "BB-UW-0114", "assessment_id": "ra_c86491074fb54213", "occurred_at": "2026-09-26T22:37:13+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5395980705" }, "output": { "decision": "approve", "score": 0.205, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-09-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.826261+00:00", "fingerprint": "52e0b66c0ebce1a8e76adbdb" }, "review": null, "ledger": { "seq": 114, "hash": "ff9d84c9fcd9bcac0e491d28572a73a4f5b5f7cc0b07e7cbfc9ad06b747b3e81", "prev_hash": "7fa43cd548fec655412206c628fbc6208d71155603b015964c1f6f9b650a327f", "recorded_at": "2026-09-26T22:40:24+00:00" } }, { "txn_id": "BB-UW-0009", "assessment_id": "ra_02a98f7166c24b0a", "occurred_at": "2026-09-27T23:29:40+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_85cdb54088" }, "output": { "decision": "approve", "score": 0.327, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-27, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.545184+00:00", "fingerprint": "683892f19487a4b529f4b693" }, "review": null, "ledger": { "seq": 9, "hash": "6ed3f6a18c2c36c6be48d7c776864f1a5c6ba512e114d6051ccc0a7077f7ead3", "prev_hash": "6d1537632160e3e4fee2cd83946165fd7452024b11183bdadbfa6bef7b760512", "recorded_at": "2026-09-27T23:31:37+00:00" } }, { "txn_id": "BB-UW-0107", "assessment_id": "ra_781f9dc25f1449eb", "occurred_at": "2026-09-29T01:19:45+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e06c4c9546" }, "output": { "decision": "approve", "score": 0.24, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.800452+00:00", "fingerprint": "ac8b8d921219e678f34b5ab6" }, "review": null, "ledger": { "seq": 107, "hash": "8523bf668e3417abaf0fb0f8d7cd0fe7990747bdf842e9277b70d34b26d522fe", "prev_hash": "d4d75ca142b4cafd10fa5c0694b47e6be0bbd36fcd1a20550e9c5fa22e7beac8", "recorded_at": "2026-09-29T01:23:31+00:00" } }, { "txn_id": "BB-UW-0125", "assessment_id": "ra_21cd7a94ef95483b", "occurred_at": "2026-09-29T02:19:06+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_598ee874c6" }, "output": { "decision": "approve", "score": 0.128, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-29, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-29 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.846764+00:00", "fingerprint": "6861f4539d4c9ace36fb310d" }, "review": null, "ledger": { "seq": 125, "hash": "56a62b0c5ab17f5f80731f42f56abbedf869ea12f9e98d6b44e10171799c10dc", "prev_hash": "d6042b00a34bf863e11a98f8dab1c57fe5a5b16442bc1d0afa403329c8ba7975", "recorded_at": "2026-09-29T02:20:53+00:00" } }, { "txn_id": "BB-UW-0106", "assessment_id": "ra_84bf41b5cd4c4109", "occurred_at": "2026-09-29T12:55:56+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_daf354230d" }, "output": { "decision": "approve", "score": 0.443, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-29, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-29 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.796904+00:00", "fingerprint": "0ac3b3e493212a3f658182f1" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 17.1, "reviewer_role": "underwriter", "resolved_at": "2026-09-30T06:04:44.435777+00:00" }, "ledger": { "seq": 106, "hash": "d4d75ca142b4cafd10fa5c0694b47e6be0bbd36fcd1a20550e9c5fa22e7beac8", "prev_hash": "668edaa09a4050c60b565728bf161710328c4ddfe4c2a0d3a00c41bab5605fe8", "recorded_at": "2026-09-29T12:59:50+00:00" } }, { "txn_id": "BB-UW-0084", "assessment_id": "ra_d921938eefb747a8", "occurred_at": "2026-09-29T16:16:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_f3d744f400" }, "output": { "decision": "refer", "score": 0.676, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-29, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-29 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-29", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.725343+00:00", "fingerprint": "f48e61a042a03641fc24609c" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 37.2, "reviewer_role": "underwriter", "resolved_at": "2026-10-01T05:28:54.473736+00:00" }, "ledger": { "seq": 84, "hash": "68b42f32979ee6295ce825a6c0bc474bd0f1a064ae1feb76d76027fa054cc4a8", "prev_hash": "d6fc1cc83e96f74df56b388266f57107b586c5ee87dd0add79f0bf4b78bf17dd", "recorded_at": "2026-09-29T16:19:06+00:00" } }, { "txn_id": "BB-UW-0079", "assessment_id": "ra_3b48ec62dc24433e", "occurred_at": "2026-09-30T00:09:08+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_3eb0c1a87c" }, "output": { "decision": "decline", "score": 0.817, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-04T00:09:08+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-09-30, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-04 with specific reasons. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.710231+00:00", "fingerprint": "deef62a2b481eb29206261eb" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 35.4, "reviewer_role": "underwriter", "resolved_at": "2026-10-01T11:36:43.133919+00:00" }, "ledger": { "seq": 79, "hash": "bc08467b99fc115610e05d8bab48e28e97b9ca1989526e531c91b88a8f3faf37", "prev_hash": "ebc3d5fb4ced408c0b88c4d0a9ee2f1ae5d374f56eda56113c7c72b11b8ad4ad", "recorded_at": "2026-09-30T00:10:44+00:00" } }, { "txn_id": "BB-UW-0126", "assessment_id": "ra_80c1918f09994483", "occurred_at": "2026-09-30T10:32:03+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_305cf754a1" }, "output": { "decision": "approve", "score": 0.076, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-09-30, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-09-30 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-09-30", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.848630+00:00", "fingerprint": "a70048dac33771de60861b12" }, "review": null, "ledger": { "seq": 126, "hash": "de6ec6e512960f3cf63dd2f2e325b3579331e23f1f78d1859ff2820efc310fbc", "prev_hash": "56a62b0c5ab17f5f80731f42f56abbedf869ea12f9e98d6b44e10171799c10dc", "recorded_at": "2026-09-30T10:34:57+00:00" } }, { "txn_id": "BB-UW-0006", "assessment_id": "ra_6dc45e9b86f74256", "occurred_at": "2026-10-02T01:11:13+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_848350a73f" }, "output": { "decision": "approve", "score": 0.377, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-02, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.537579+00:00", "fingerprint": "9138d2affadc531b890fe248" }, "review": null, "ledger": { "seq": 6, "hash": "997247e1d1232831d722095f4efef3b3f92fb8d830baefeee20c0d13025feb8c", "prev_hash": "875aa9f6fb59c10bab07bf31c0af277c87bcf32048b882d96446ce157b641b61", "recorded_at": "2026-10-02T01:13:25+00:00" } }, { "txn_id": "BB-UW-0085", "assessment_id": "ra_c16c9c1c5b10415f", "occurred_at": "2026-10-02T11:46:06+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_4a0c092219" }, "output": { "decision": "approve", "score": 0.053, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-02, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-02 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-02", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.729118+00:00", "fingerprint": "d5ef879e861a412e684ba5e2" }, "review": null, "ledger": { "seq": 85, "hash": "524170528f96cafdd4e1ea20a0f1ea2d8e23192957980234039f1d00ba539e70", "prev_hash": "68b42f32979ee6295ce825a6c0bc474bd0f1a064ae1feb76d76027fa054cc4a8", "recorded_at": "2026-10-02T11:49:13+00:00" } }, { "txn_id": "BB-UW-0008", "assessment_id": "ra_093b8c1f8aff48fb", "occurred_at": "2026-10-06T11:01:18+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7ab2c55523" }, "output": { "decision": "approve", "score": 0.383, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-06, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-06 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.543723+00:00", "fingerprint": "627ade718bf2481c559e3ddb" }, "review": null, "ledger": { "seq": 8, "hash": "6d1537632160e3e4fee2cd83946165fd7452024b11183bdadbfa6bef7b760512", "prev_hash": "7ac49466b3155551da453e69d5002d5a7d102b21f36ac555280db56724d66549", "recorded_at": "2026-10-06T11:04:28+00:00" } }, { "txn_id": "TX-RNW-0199", "assessment_id": "ra_fa4bc92caaad4f2e", "occurred_at": "2026-10-06T15:00:00+00:00", "state": "TX", "line_of_business": "home", "stage": "renewal", "program_area": "Ratemaking/Rate Classification/Schedule Rating/Premium Audits", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "aerial_imagery", "property", "claims_history" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": false, "credit_used": false, "data_source_refs": [ "imagery://vendor/roof-condition" ], "consumer_token": "tok_showcase_0199" }, "output": { "decision": "non_renew", "score": 0.84, "reason_codes": [ "ROOF-3" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": null, "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": "Homeowners non-renewal driven by an aerial-imagery roof-condition score; no notice generated; no human review.", "assessment": { "disposition": "FLAG", "summary": "FLAG: 3 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-06, uw-triage v2.4 (an internally developed system) produced 'non renew' in a TX home renewal transaction. Inputs included aerial imagery, property, claims history. No human reviewed the output before it took effect. The outcome was adverse to the consumer; no notice to the consumer is recorded. RegBench applied 11 rule(s) in force on 2026-10-06 and recorded 3 finding(s); the most serious is 'Consequential AI-supported decision taken without a person reviewing and agreeing' (high, L4).", "findings": [ { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "precedent.ai-flagged-cancellation-without-notice", "layer": "L5", "judgment_basis": "precedent", "advisory": true, "severity": "medium", "title": "AI-flagged non-renewal or decline with no notice recorded (pattern of the 2026 Pennsylvania agreement)", "required_action": "Confirm the notice and document deadlines before the cancellation or decline takes effect; record the notice on the transaction.", "exhibits": [ "C" ], "citations": [ { "provision_id": "precedent-pa-geico-2026", "label": "Pennsylvania Attorney General agreement with GEICO (announced May 2026) \u2014 AI-enabled underwriting review and cancellation without adequate notice", "source_url": "https://www.attorneygeneral.gov/?p=33518", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-06", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:25.371784+00:00", "fingerprint": "bb5cdcc31dd03749dad0eef9" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 10.9, "reviewer_role": "compliance", "resolved_at": "2026-10-07T01:56:23.646212+00:00" }, "ledger": { "seq": 1128, "hash": "5de84fec1984ed9a0fe4732b619be4eb5f8bb05be6a9509823ce61cb53d223b1", "prev_hash": "d66a8c916be6b32ef077524a3996446f4fec69c3e554652db5876341e838de18", "recorded_at": "2026-10-06T15:01:29+00:00" } }, { "txn_id": "BB-UW-0061", "assessment_id": "ra_acfb00d3d0854914", "occurred_at": "2026-10-07T23:04:01+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_85f975cc37" }, "output": { "decision": "approve", "score": 0.194, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-07, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-07 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-07", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.665796+00:00", "fingerprint": "593085beb9aaff06f3abc0f1" }, "review": null, "ledger": { "seq": 61, "hash": "f86b262c1ef79c94a8b45f55a2314c8e9d60e31dfe9299412f620cfc36e29332", "prev_hash": "0740718325691e0ca44b6e793f743da597f6bc0bb1f576aafd997261f71b1174", "recorded_at": "2026-10-07T23:04:53+00:00" } }, { "txn_id": "BB-UW-0001", "assessment_id": "ra_fc4f18c392c24d37", "occurred_at": "2026-10-08T09:16:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_dca5aec797" }, "output": { "decision": "decline", "score": 0.874, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-14T09:16:00+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-08, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-14 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-08 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.528150+00:00", "fingerprint": "557a70b92d4ea3a503a0fcbc" }, "review": { "status": "resolved", "outcome": "overridden", "hours_to_resolve": 5.7, "reviewer_role": "underwriter", "resolved_at": "2026-10-08T15:03:13.658973+00:00" }, "ledger": { "seq": 1, "hash": "b521b1071255d2423c030b9dc80f7c2712b6755e366ce596e51e1e13646afbeb", "prev_hash": "0000000000000000000000000000000000000000000000000000000000000000", "recorded_at": "2026-10-08T09:19:12+00:00" } }, { "txn_id": "BB-UW-0036", "assessment_id": "ra_10435299d75340ba", "occurred_at": "2026-10-08T19:37:52+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_43d1e29034" }, "output": { "decision": "approve", "score": 0.073, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-08, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-08 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-08", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.613091+00:00", "fingerprint": "bf4728126439c759ac803abf" }, "review": null, "ledger": { "seq": 36, "hash": "2660771fe0ef1edbe35350bb3af7d8cd69fa5eddb5ed8565a2846595848af2b2", "prev_hash": "b0f71cf264db7639805e08e5a73446a0dbeb75f35bee947461dfc81822717ab2", "recorded_at": "2026-10-08T19:40:26+00:00" } }, { "txn_id": "BB-UW-0011", "assessment_id": "ra_f406c8a619ab41e3", "occurred_at": "2026-10-09T00:52:54+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_734242c225" }, "output": { "decision": "approve", "score": 0.341, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.548127+00:00", "fingerprint": "62e9144721bf4b2012d7f541" }, "review": null, "ledger": { "seq": 11, "hash": "37fb57ef1172d4dc84c7f5fc9b9114a42622f23fcaa10d2fb807a9af689b31fe", "prev_hash": "3d723048b894580fd6a2e1ebe9f63e5f45993f484e0ca19b73cecd70c4ed8500", "recorded_at": "2026-10-09T00:55:31+00:00" } }, { "txn_id": "BB-UW-0104", "assessment_id": "ra_b82748b48bc846fd", "occurred_at": "2026-10-09T04:23:49+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_e1fe5961f2" }, "output": { "decision": "approve", "score": 0.409, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.785948+00:00", "fingerprint": "f7e66de5ef3b92841201d3bc" }, "review": null, "ledger": { "seq": 104, "hash": "1adf359a02c5e05aa106eabb00c16c0023f9ea8562951ef5de3c40f49b325bfd", "prev_hash": "7cd1722422bd37b40d84f5a7a86e4e1c2a59c46cc06e71f07304e463586f5d0c", "recorded_at": "2026-10-09T04:27:11+00:00" } }, { "txn_id": "BB-UW-0021", "assessment_id": "ra_c91cd066d60048e4", "occurred_at": "2026-10-09T08:42:28+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_f4e50588a5" }, "output": { "decision": "approve", "score": 0.159, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.576853+00:00", "fingerprint": "2db520a46b3bceb7277a58ef" }, "review": null, "ledger": { "seq": 21, "hash": "cf76eafe0f1f806378b08932609a7546ccce875bfeb19188c7cb9277bd19150e", "prev_hash": "bb4c30cb67dda82960cad0dab6c491c10c1b7e86c3c9da2c920ff91d874128aa", "recorded_at": "2026-10-09T08:45:00+00:00" } }, { "txn_id": "BB-UW-0144", "assessment_id": "ra_490b771e788f4b82", "occurred_at": "2026-10-09T13:03:31+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_b7c02928d8" }, "output": { "decision": "approve", "score": 0.156, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-09, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-09 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-09", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.886827+00:00", "fingerprint": "4c4643b7152bd30c41efd913" }, "review": null, "ledger": { "seq": 144, "hash": "7f41bb1cecbcea5c283820b3f327f2ea1646461f1b7fe1fedcf2d5d2b9a68689", "prev_hash": "59e5f414b94246103c63d479721d772d206c9706418448b48ce7b615b6d4065b", "recorded_at": "2026-10-09T13:06:09+00:00" } }, { "txn_id": "BB-UW-0080", "assessment_id": "ra_ef44910a67d5484d", "occurred_at": "2026-10-11T21:34:42+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7f042f7867" }, "output": { "decision": "approve", "score": 0.236, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-11, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-11 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-11", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.714129+00:00", "fingerprint": "cbe1547e6b69071c958aa272" }, "review": null, "ledger": { "seq": 80, "hash": "b333ef195377342083269ad98c0463758d0e6586a76ecf2416fd52b41d97c766", "prev_hash": "bc08467b99fc115610e05d8bab48e28e97b9ca1989526e531c91b88a8f3faf37", "recorded_at": "2026-10-11T21:36:39+00:00" } }, { "txn_id": "BB-UW-0127", "assessment_id": "ra_97a3f6bce55a42cc", "occurred_at": "2026-10-12T18:14:03+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_fd0c996f04" }, "output": { "decision": "approve", "score": 0.16, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-12, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-12 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.850070+00:00", "fingerprint": "19a49de5f2ae33a70957d1f3" }, "review": null, "ledger": { "seq": 127, "hash": "9fe4d23cc6d8c5e978ef2d1bb2a481b149a087b578fa2fabeeb1c2c2f1ddebca", "prev_hash": "de6ec6e512960f3cf63dd2f2e325b3579331e23f1f78d1859ff2820efc310fbc", "recorded_at": "2026-10-12T18:17:30+00:00" } }, { "txn_id": "BB-UW-0101", "assessment_id": "ra_395ffb0faaa14d4e", "occurred_at": "2026-10-12T20:41:15+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_d4e1bf119c" }, "output": { "decision": "decline", "score": 0.933, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-15T20:41:15+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-12, uw-triage v2.4 (an internally developed system) produced 'decline' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-15 with specific reasons. RegBench applied 10 rule(s) in force on 2026-10-12 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-12", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.774145+00:00", "fingerprint": "5698e07bc7e18e9e8933b092" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 21.5, "reviewer_role": "underwriter", "resolved_at": "2026-10-13T18:11:50.188949+00:00" }, "ledger": { "seq": 101, "hash": "19ad4e4907b466dfa89302f042f1495f6d24cd8651977cc31d806420c3afbc41", "prev_hash": "b449e320614fdcf635b043919db3980bbea6a94c33366158c9e6b6e8aab19d91", "recorded_at": "2026-10-12T20:42:42+00:00" } }, { "txn_id": "BB-UW-0070", "assessment_id": "ra_fb7939a32f70442c", "occurred_at": "2026-10-13T01:17:46+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_defdeac4b1" }, "output": { "decision": "decline", "score": 0.897, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-17T01:17:46+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 4 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-13, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was adverse to the consumer; a notice was sent on 2026-10-17 with reasons stated. RegBench applied 11 rule(s) in force on 2026-10-13 and recorded 4 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "tx.tdi.consequential-decision-without-human-review", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "high", "title": "Consequential AI-supported decision taken without a person reviewing and agreeing", "required_action": "Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "tx-tdi-b-0003-26-human-review", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026), Use of Artificial Intelligence \u2014 human review of consequential decisions", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true }, { "provision_id": "tx-tdi-b-0003-26-governance", "label": "TDI Commissioner's Bulletin B-0003-26 (12 June 2026) \u2014 governance, procedures on request, and monitoring", "source_url": "https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html", "verified": true } ] }, { "rule_id": "doctrine.unreviewed-adverse-high-risk", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Adverse outcome from a high-risk AI system with no human review", "required_action": "Route to the carrier's reviewer; record the review and any override on the transaction.", "exhibits": [ "B", "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.685648+00:00", "fingerprint": "fb1ff0ac3e37345fb18df130" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 3.9, "reviewer_role": "underwriter", "resolved_at": "2026-10-13T05:11:22.473680+00:00" }, "ledger": { "seq": 70, "hash": "52dccfe65eedb37f144748b78e6f9729c7119a39b5b37e4b153417a24cf443d1", "prev_hash": "ab0eda9df0814e3a92d0119d45255f30107bcb6346132e45a6ffc2183a0447b0", "recorded_at": "2026-10-13T01:20:09+00:00" } }, { "txn_id": "BB-UW-0004", "assessment_id": "ra_7731bf1099a64c46", "occurred_at": "2026-10-13T11:33:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_494ac87fc0" }, "output": { "decision": "approve", "score": 0.281, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-13, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-13 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-13", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.533273+00:00", "fingerprint": "722ff40d8832c01f59a9f40d" }, "review": null, "ledger": { "seq": 4, "hash": "abc560a63897b1057ff0140eac16b9202e5b370c45472d0f2f33fb74243431cd", "prev_hash": "38607c1d1af3bcf9d3628ad652b3c3aa5f384d9a6580ce45d18b3a941a82c6da", "recorded_at": "2026-10-13T11:36:36+00:00" } }, { "txn_id": "BB-UW-0077", "assessment_id": "ra_0708422ee1544166", "occurred_at": "2026-10-14T00:13:49+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_27059f963d" }, "output": { "decision": "approve", "score": 0.087, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.700732+00:00", "fingerprint": "dd32396189f45bbc4b2b51b1" }, "review": null, "ledger": { "seq": 77, "hash": "0f5090f4f191b7cef2a61574f4002d057e71a5ab106cce40c57a48c71846e945", "prev_hash": "09018bc6253bd48f3ad088c53c374e760f7825586461fb0414ac7ebdc338feab", "recorded_at": "2026-10-14T00:17:27+00:00" } }, { "txn_id": "BB-UW-0035", "assessment_id": "ra_28b17a10d77e48c6", "occurred_at": "2026-10-14T01:56:30+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_6a7a957971" }, "output": { "decision": "refer", "score": 0.598, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, uw-triage v2.4 (an internally developed system) produced 'refer' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.610485+00:00", "fingerprint": "f21ba2f8baea3b893307bb78" }, "review": null, "ledger": { "seq": 35, "hash": "b0f71cf264db7639805e08e5a73446a0dbeb75f35bee947461dfc81822717ab2", "prev_hash": "0d3c86899cc2fb5224745959e5d7faaf5f1cb19d7430d40a89dd7293fb9c25e2", "recorded_at": "2026-10-14T01:58:40+00:00" } }, { "txn_id": "BB-UW-0034", "assessment_id": "ra_39ab39d352c24e4c", "occurred_at": "2026-10-14T10:13:24+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a907465834" }, "output": { "decision": "approve", "score": 0.406, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-14, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-14 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-14", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.605762+00:00", "fingerprint": "4821223ef67bf96d94e4a412" }, "review": null, "ledger": { "seq": 34, "hash": "0d3c86899cc2fb5224745959e5d7faaf5f1cb19d7430d40a89dd7293fb9c25e2", "prev_hash": "e34e2d531f4aa0b9c5f02d614dd38bc7e2580f89defb602ca635a307f92beb3e", "recorded_at": "2026-10-14T10:17:24+00:00" } }, { "txn_id": "BB-UW-0053", "assessment_id": "ra_0d34bc9a53cb4ce1", "occurred_at": "2026-10-15T02:27:38+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_358445d1a4" }, "output": { "decision": "approve", "score": 0.21, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-15, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.649678+00:00", "fingerprint": "a3a301c7cabcca241d641f8c" }, "review": null, "ledger": { "seq": 53, "hash": "27c407513956bb9bd8c14393eb4a6592d191f2df1d5dca580c4eabadeb55e2c6", "prev_hash": "b87b3d77cf9a11f79cb9d7b5796ac8db0d7f80d85af01922f8cb3da954f73106", "recorded_at": "2026-10-15T02:29:17+00:00" } }, { "txn_id": "BB-UW-0098", "assessment_id": "ra_01f328499b2146e0", "occurred_at": "2026-10-15T14:53:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_e94d8984a3" }, "output": { "decision": "approve", "score": 0.189, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-15, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-15 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.765311+00:00", "fingerprint": "6ab443c05c6fe5bd4895828f" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 71.3, "reviewer_role": "underwriter", "resolved_at": "2026-10-18T14:15:31.686314+00:00" }, "ledger": { "seq": 98, "hash": "bf3ac86a5a03e810e044519e0b49dbbf2d921c951a43e43bb41f6cd778709547", "prev_hash": "7ef37006fe2e18bb3d431351d05f7675ae29bc9ea9adf62201fb618b8d4772c4", "recorded_at": "2026-10-15T14:56:21+00:00" } }, { "txn_id": "TX-UW-0311", "assessment_id": "ra_92dbfb02417f4df0", "occurred_at": "2026-10-15T15:00:00+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_showcase_0311" }, "output": { "decision": "approve", "score": 0.21, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": "Standard new-business approval with a consumer report and underwriter sign-off.", "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-15, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-15 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-15", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:25.365522+00:00", "fingerprint": "f63f87a4e8ca9237edf5ba00" }, "review": null, "ledger": { "seq": 1126, "hash": "d66a8c916be6b32ef077524a3996446f4fec69c3e554652db5876341e838de18", "prev_hash": "852d7d9347da922a27f5afce10ff456ee67fe02425498f4426a8456d1b223d44", "recorded_at": "2026-10-15T15:01:48+00:00" } }, { "txn_id": "BB-UW-0026", "assessment_id": "ra_f64bebe546554592", "occurred_at": "2026-10-16T21:54:32+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_1921f34390" }, "output": { "decision": "approve", "score": 0.092, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-16, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-16 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-16", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.586098+00:00", "fingerprint": "84be5c744f09133d3569d33d" }, "review": null, "ledger": { "seq": 26, "hash": "8761189c2723dc846aa47e7904f92a10ed6392ba4f034fbd579ad628c218dc02", "prev_hash": "b17a444da61df1db539a7b9a3ee1468a001e658a5110ed9b0ceada7ff378a923", "recorded_at": "2026-10-16T21:58:21+00:00" } }, { "txn_id": "BB-UW-0049", "assessment_id": "ra_761b234c3939483d", "occurred_at": "2026-10-18T01:08:02+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_5254eb1e9f" }, "output": { "decision": "decline", "score": 0.813, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-20T01:08:02+00:00", "reasons_provided": true, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "HOLD", "summary": "HOLD: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Held for the carrier's compliance review before the outcome is communicated; the carrier decides.", "narrative": "On 2026-10-18, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-20 with reasons stated. RegBench applied 11 rule(s) in force on 2026-10-18 and recorded 2 finding(s); the most serious is 'Credit-based adverse effect without the \u00a7 559.054 notice of reasons' (high, L2).", "findings": [ { "rule_id": "tx.559.credit-adverse-effect-notice", "layer": "L2", "judgment_basis": "black_letter", "advisory": false, "severity": "high", "title": "Credit-based adverse effect without the \u00a7 559.054 notice of reasons", "required_action": "Hold until a notice stating the primary credit factors is generated and recorded within the 30-day window.", "exhibits": [ "C" ], "citations": [ { "provision_id": "tx-ins-559-054", "label": "Tex. Ins. Code \u00a7 559.054 \u2014 notice of action resulting in adverse effect (credit report)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true }, { "provision_id": "tx-ins-559-001", "label": "Tex. Ins. Code \u00a7 559.001(1) \u2014 'adverse effect' (credit information in personal lines)", "source_url": "https://statutes.capitol.texas.gov/?tab=1&code=IN&chapter=IN.559", "verified": true } ] }, { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-18", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.643138+00:00", "fingerprint": "fd9ff826ef704765a3a3346a" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 4.5, "reviewer_role": "underwriter", "resolved_at": "2026-10-18T05:40:48.912606+00:00" }, "ledger": { "seq": 49, "hash": "fd55d10e820431438117056d3ecf300a50d405ca9ec6e11c94f142284305895b", "prev_hash": "260aa859521fc49eddcfcb70a85518622bc5d300058c0e85b76980081c667f6e", "recorded_at": "2026-10-18T01:10:04+00:00" } }, { "txn_id": "BB-UW-0025", "assessment_id": "ra_a1aa81777013470b", "occurred_at": "2026-10-20T03:15:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_89cab39a07" }, "output": { "decision": "approve", "score": 0.35, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-20, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.583901+00:00", "fingerprint": "93a216b1786501563ec4071e" }, "review": null, "ledger": { "seq": 25, "hash": "b17a444da61df1db539a7b9a3ee1468a001e658a5110ed9b0ceada7ff378a923", "prev_hash": "1fd83882254587608e93f17509fef55156720474276d280ffa671f86d6630bac", "recorded_at": "2026-10-20T03:19:22+00:00" } }, { "txn_id": "BB-UW-0132", "assessment_id": "ra_5a63ff062a8e46d5", "occurred_at": "2026-10-20T19:27:36+00:00", "state": "TX", "line_of_business": "home", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7eb79542a7" }, "output": { "decision": "approve", "score": 0.3, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-20, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX home underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-20 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-20", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.862267+00:00", "fingerprint": "475c2862664636c716d16222" }, "review": null, "ledger": { "seq": 132, "hash": "d5bad2ac3373823956d2dd21ad68382aeb15e7ca898cd92dbeb8e01b51790d0b", "prev_hash": "44d826bbfce7ae2d012744c6a08455775c2ba4a4ce9d94185282de7ec42357d4", "recorded_at": "2026-10-20T19:31:11+00:00" } }, { "txn_id": "BB-UW-0016", "assessment_id": "ra_f02f1710a14e4d0b", "occurred_at": "2026-10-22T19:05:38+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_814e564cbf" }, "output": { "decision": "approve", "score": 0.129, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-22, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-22 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-22", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.562386+00:00", "fingerprint": "cab6281ec0074e0379a96b28" }, "review": null, "ledger": { "seq": 16, "hash": "f51a072b17b97b8ebca62fe8457825f3e9c558237af7b3ae32c5c1aee547d55d", "prev_hash": "ef47edfe70255c3bbab2fefe6108428405a46af427aaac11fe3b9e4cc4efa9d8", "recorded_at": "2026-10-22T19:09:11+00:00" } }, { "txn_id": "BB-UW-0050", "assessment_id": "ra_b2489f9d3c6642d8", "occurred_at": "2026-10-23T12:57:43+00:00", "state": "CO", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": true, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_3661b25908" }, "output": { "decision": "approve", "score": 0.101, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a CO auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. External consumer data and information sources were present. A consumer report was used. A human reviewed the output by a underwriter. The outcome was not adverse to the consumer. RegBench applied 10 rule(s) in force on 2026-10-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.644887+00:00", "fingerprint": "3c692f7db89d3931f261a3d9" }, "review": null, "ledger": { "seq": 50, "hash": "b37f14ca62223a385d4a8f06d9069e46cb56db1fcb2f1a7339bd5de461b7b244", "prev_hash": "fd55d10e820431438117056d3ecf300a50d405ca9ec6e11c94f142284305895b", "recorded_at": "2026-10-23T12:58:51+00:00" } }, { "txn_id": "BB-UW-0076", "assessment_id": "ra_c412d14d76624ef5", "occurred_at": "2026-10-23T14:34:15+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_110ab83474" }, "output": { "decision": "approve", "score": 0.37, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-23, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-23 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.699268+00:00", "fingerprint": "b73d4759d80b89472129001e" }, "review": null, "ledger": { "seq": 76, "hash": "09018bc6253bd48f3ad088c53c374e760f7825586461fb0414ac7ebdc338feab", "prev_hash": "d9e66d632853bfc5c62aa6fede7fe03538999d9fd75eaac686656a86a5238c44", "recorded_at": "2026-10-23T14:37:30+00:00" } }, { "txn_id": "BB-UW-0091", "assessment_id": "ra_1a2d3d7079bb480a", "occurred_at": "2026-10-23T19:28:22+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_a605611673" }, "output": { "decision": "decline", "score": 0.964, "reason_codes": [ "R12", "R31" ], "premium_delta_pct": null }, "human_involvement": { "reviewed": true, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": true, "notice_sent_at": "2026-10-28T19:28:22+00:00", "reasons_provided": true, "specific_reasons": true, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 1 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-23, uw-triage v2.4 (an internally developed system) produced 'decline' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. A human reviewed the output by a underwriter. The outcome was adverse to the consumer; a notice was sent on 2026-10-28 with specific reasons. RegBench applied 11 rule(s) in force on 2026-10-23 and recorded 1 finding(s); the most serious is 'Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)' (medium, L6).", "findings": [ { "rule_id": "doctrine.zip-credit-interaction-proxy", "layer": "L6", "judgment_basis": "doctrine", "advisory": true, "severity": "medium", "title": "Geography and credit used together in an adverse decision (proxy pattern to test at portfolio level)", "required_action": "Confirm the system's portfolio outcome testing covers the geography-credit interaction; link the result.", "exhibits": [ "C" ], "citations": [ { "provision_id": "naic-ai-bulletin-s3-1-ais-program", "label": "NAIC Model Bulletin on the Use of AI Systems by Insurers (adopted 4 Dec 2023), Section 3, 1.0 General Guidelines \u2014 the written AIS Program", "source_url": "https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-23", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.752156+00:00", "fingerprint": "2e721a05917cbd7825c2af94" }, "review": { "status": "resolved", "outcome": "remediated", "hours_to_resolve": 58.8, "reviewer_role": "underwriter", "resolved_at": "2026-10-26T06:19:12.868202+00:00" }, "ledger": { "seq": 91, "hash": "6bbecd1e12300ff8a89decfa92af2b1d4b823cb62874fed65bbff494003507bc", "prev_hash": "2a2d9076d74636ee742243252f24775dbce40cf65d0808a885bf9210d1cc3639", "recorded_at": "2026-10-23T19:31:42+00:00" } }, { "txn_id": "BB-UW-0046", "assessment_id": "ra_e582bb1de65d4ab1", "occurred_at": "2026-10-25T09:51:20+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [], "consumer_token": "tok_3f9c2a2dc0" }, "output": { "decision": "approve", "score": 0.4, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "FLAG", "summary": "FLAG: 2 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. Routed to the carrier's compliance reviewer with cited findings.", "narrative": "On 2026-10-25, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-25 and recorded 2 finding(s); the most serious is 'Consumer report used with no data source reference recorded' (low, L0).", "findings": [ { "rule_id": "us.fcra.permissible-purpose-undocumented", "layer": "L0", "judgment_basis": "black_letter", "advisory": false, "severity": "low", "title": "Consumer report used with no data source reference recorded", "required_action": "Record the consumer reporting agency and report reference on the transaction.", "exhibits": [ "C", "D" ], "citations": [ { "provision_id": "us-fcra-1681b-a-3-c", "label": "15 U.S.C. \u00a7 1681b(a)(3)(C) \u2014 permissible purpose: underwriting of insurance", "source_url": "https://www.law.cornell.edu/uscode/text/15/1681b", "verified": true } ] }, { "rule_id": "tx.tdi.third-party-data-source-unrecorded", "layer": "L4", "judgment_basis": "guidance", "advisory": false, "severity": "low", "title": "Third-party data used with no source recorded", "required_action": "Record the data source reference so the data can be traced and corrected.", "exhibits": [ "D" ], "citations": [ { "provision_id": "tx-tdi-b-0036-20", "label": "TDI Commissioner's Bulletin B-0036-20 (30 Sept 2020) \u2014 insurers' use of third-party data", "source_url": "https://www.tdi.texas.gov/bulletins/2020/B-0036-20.html", "verified": true } ] } ], "point_in_time": { "as_of": "2026-10-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.635682+00:00", "fingerprint": "7e5ced02bdc7c0d85bcc32f2" }, "review": { "status": "resolved", "outcome": "confirmed-no-change", "hours_to_resolve": 26.5, "reviewer_role": "underwriter", "resolved_at": "2026-10-26T12:26:13.907982+00:00" }, "ledger": { "seq": 46, "hash": "e3332ee401cddcc0f865ef0e8eeaca28ebfa72b533746392ccb54b0a22f11951", "prev_hash": "8a9cdc4431d7c83b488b1de7997065a4b3d3b59f98715a5fed6f736764f1bf44", "recorded_at": "2026-10-25T09:55:03+00:00" } }, { "txn_id": "BB-UW-0137", "assessment_id": "ra_63b901ea3beb4e8c", "occurred_at": "2026-10-25T22:13:28+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2d4d32a879" }, "output": { "decision": "approve", "score": 0.224, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-25, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-25 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-25", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.870545+00:00", "fingerprint": "1555fb18191e7e84138aa453" }, "review": null, "ledger": { "seq": 137, "hash": "f703a643c649bbf1a9e771f1d8aaee2f496367febd7c1fea3ee53df38fa5adbe", "prev_hash": "4e61ecffcd51833d52b9b7a2c79a04e30d3d10419093a9a1876034b980e9c1fa", "recorded_at": "2026-10-25T22:17:25+00:00" } }, { "txn_id": "BB-UW-0044", "assessment_id": "ra_c05ab96ea8d446c8", "occurred_at": "2026-10-26T15:56:13+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_7aa5d50dd0" }, "output": { "decision": "approve", "score": 0.361, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.631378+00:00", "fingerprint": "e83ded6f6d08f49f97242039" }, "review": null, "ledger": { "seq": 44, "hash": "a4d8638aec9c14b6f5dffa688655402843599c3396806fda6fa8e4bf72c1988b", "prev_hash": "1e8d31db99b3bd7f1dee53493c296370d1f366adc5200bbf640779eb4ff21b3b", "recorded_at": "2026-10-26T15:56:55+00:00" } }, { "txn_id": "BB-UW-0057", "assessment_id": "ra_9f71bf4ceff84e52", "occurred_at": "2026-10-26T16:32:49+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_2bac531842" }, "output": { "decision": "approve", "score": 0.259, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.659018+00:00", "fingerprint": "0dd8c2c303542087b2716738" }, "review": null, "ledger": { "seq": 57, "hash": "1e9e6c6424ea106167b9f01871307ba3edbdbf2feea8472ab1c24f442b8f1304", "prev_hash": "09bcbd50ffebfa12ef23a4ec7447bee628ae4f2620045ea1b60b5f8fa2a96fa0", "recorded_at": "2026-10-26T16:33:34+00:00" } }, { "txn_id": "BB-UW-0109", "assessment_id": "ra_97eac5920b0b412e", "occurred_at": "2026-10-26T17:56:48+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_4eca8c799f" }, "output": { "decision": "approve", "score": 0.328, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-26, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-26 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-26", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.808672+00:00", "fingerprint": "0dd8c2c303542087b2716738" }, "review": null, "ledger": { "seq": 109, "hash": "24633efa909dfd177e3a043b086462d39c7f07836a4f85631c8d56f1b69c7e69", "prev_hash": "9039f67c13b0328e973cc3948d76f2c97c27054d122b78be615d248d69e52888", "recorded_at": "2026-10-26T17:57:31+00:00" } }, { "txn_id": "BB-UW-0100", "assessment_id": "ra_0cacc4a90bd74268", "occurred_at": "2026-10-27T19:21:41+00:00", "state": "TX", "line_of_business": "auto", "stage": "underwriting", "program_area": "Underwriting/Eligibility", "ai_system": { "system_id": "uw-triage", "model_version": "2.4", "vendor": "internal", "vendor_name": null, "high_risk_class": "underwriting" }, "inputs": { "feature_classes": [ "application_data", "claims_history", "consumer_report", "credit", "geography" ], "protected_factors_used": [], "ecdis_present": false, "consumer_report_used": true, "credit_used": true, "data_source_refs": [ "cra://demo-bureau/report" ], "consumer_token": "tok_bc4e9fe834" }, "output": { "decision": "approve", "score": 0.169, "reason_codes": [], "premium_delta_pct": null }, "human_involvement": { "reviewed": false, "reviewer_role": "underwriter", "overridden": false }, "adverse_action": { "is_adverse": false, "notice_sent_at": null, "reasons_provided": false, "specific_reasons": false, "ecdis_disclosed": false, "appeal_channel": null }, "carrier_note": null, "assessment": { "disposition": "PASS", "summary": "PASS: 0 finding(s) under rules 2026.10+dbe5ddf6, corpus 2026-10-30-5f07c521. No regulatory finding; the case file is recorded.", "narrative": "On 2026-10-27, uw-triage v2.4 (an internally developed system) produced 'approve' in a TX auto underwriting transaction. Inputs included application data, claims history, consumer report, credit, geography. A consumer report was used. No human reviewed the output before it took effect. The outcome was not adverse to the consumer. RegBench applied 11 rule(s) in force on 2026-10-27 and recorded no findings.", "findings": [], "point_in_time": { "as_of": "2026-10-27", "rules_version": "2026.10+dbe5ddf6", "corpus_snapshot": "2026-10-30-5f07c521", "engine_version": "regbench-engine/0.4.0", "model_version": "deterministic-1" }, "evaluated_at": "2026-10-30T11:23:22.769163+00:00", "fingerprint": "03976e05118e58c5d64499d7" }, "review": null, "ledger": { "seq": 100, "hash": "b449e320614fdcf635b043919db3980bbea6a94c33366158c9e6b6e8aab19d91", "prev_hash": "989887f541d517f6a05da1f110b0788af8887394fec52609f8793fd59add5cab", "recorded_at": "2026-10-27T19:23:51+00:00" } } ] } ] }, "exhibit_d": { "title": "Exhibit D: AI Model Data", "by_line_of_business": [ { "line_of_business": "auto", "models": [ "agent-assist", "claims-fastpath", "fnol-fraud", "price-opt", "uw-triage" ], "rows": [ { "ref": 1, "data_element": "Aerial Imagery", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 2, "data_element": "Age, Gender, Ethnicity/Race", "used": false }, { "ref": 3, "data_element": "Consumer or Other Type of Insurance/Risk Score", "used": true, "models": [ "price-opt", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 4, "data_element": "Crime Statistics", "used": false }, { "ref": 5, "data_element": "Criminal Convictions (Exclude Auto-Related Convictions)", "used": false }, { "ref": 6, "data_element": "Driving Behavior", "used": true, "models": [ "price-opt" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 7, "data_element": "Education Level (Including school aptitude scores, etc.)", "used": false }, { "ref": 8, "data_element": "Facial or Body Detection / Recognition / Analysis", "used": false }, { "ref": 9, "data_element": "Geocoding (including address, city, county, state, ZIP code, lat/long, MSA/CSA, etc.)", "used": true, "models": [ "fnol-fraud", "price-opt", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 10, "data_element": "Geo-Demographics (including ZIP/county-based demographic characteristics)", "used": false }, { "ref": 11, "data_element": "Household Composition", "used": false }, { "ref": 12, "data_element": "Image/video Analysis", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 13, "data_element": "Income", "used": false }, { "ref": 14, "data_element": "Job History", "used": false }, { "ref": 15, "data_element": "Loss Experience", "used": true, "models": [ "claims-fastpath", "fnol-fraud", "price-opt", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 16, "data_element": "Medical, including Biometrics, genetic information, pre-existing conditions, diagnostic data, etc.", "used": false }, { "ref": 17, "data_element": "Natural Catastrophe Hazard (Fire, Wind, Hail, Earthquake, Severe Convective Storms)", "used": false }, { "ref": 18, "data_element": "Online social media, including characteristics for targeted advertising", "used": true, "models": [ "fnol-fraud" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": false, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 19, "data_element": "Personal Financial Information", "used": true, "models": [ "price-opt", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 20, "data_element": "Reasonable Accommodations or Policy Modifications (granted or requested)", "used": false }, { "ref": 21, "data_element": "Telematics/Usage-based insurance", "used": true, "models": [ "price-opt" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 22, "data_element": "Vehicle-Specific Data including VIN characteristics", "used": true, "models": [ "price-opt" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 23, "data_element": "Voice Analysis", "used": false }, { "ref": 24, "data_element": "Weather", "used": false }, { "ref": 25, "data_element": "Other: Non-Traditional Data Elements (Regulator to specify)", "used": true, "models": [ "claims-fastpath", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] } ] }, { "line_of_business": "home", "models": [ "claims-fastpath", "fnol-fraud", "uw-triage" ], "rows": [ { "ref": 1, "data_element": "Aerial Imagery", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 2, "data_element": "Age, Gender, Ethnicity/Race", "used": false }, { "ref": 3, "data_element": "Consumer or Other Type of Insurance/Risk Score", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 4, "data_element": "Crime Statistics", "used": false }, { "ref": 5, "data_element": "Criminal Convictions (Exclude Auto-Related Convictions)", "used": false }, { "ref": 6, "data_element": "Driving Behavior", "used": false }, { "ref": 7, "data_element": "Education Level (Including school aptitude scores, etc.)", "used": false }, { "ref": 8, "data_element": "Facial or Body Detection / Recognition / Analysis", "used": false }, { "ref": 9, "data_element": "Geocoding (including address, city, county, state, ZIP code, lat/long, MSA/CSA, etc.)", "used": true, "models": [ "fnol-fraud", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 10, "data_element": "Geo-Demographics (including ZIP/county-based demographic characteristics)", "used": false }, { "ref": 11, "data_element": "Household Composition", "used": false }, { "ref": 12, "data_element": "Image/video Analysis", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 13, "data_element": "Income", "used": false }, { "ref": 14, "data_element": "Job History", "used": false }, { "ref": 15, "data_element": "Loss Experience", "used": true, "models": [ "claims-fastpath", "fnol-fraud", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 16, "data_element": "Medical, including Biometrics, genetic information, pre-existing conditions, diagnostic data, etc.", "used": false }, { "ref": 17, "data_element": "Natural Catastrophe Hazard (Fire, Wind, Hail, Earthquake, Severe Convective Storms)", "used": false }, { "ref": 18, "data_element": "Online social media, including characteristics for targeted advertising", "used": true, "models": [ "fnol-fraud" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": false, "third_party_vendors": [ "Sentinel Analytics (fictional)" ] }, { "ref": 19, "data_element": "Personal Financial Information", "used": true, "models": [ "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] }, { "ref": 20, "data_element": "Reasonable Accommodations or Policy Modifications (granted or requested)", "used": false }, { "ref": 21, "data_element": "Telematics/Usage-based insurance", "used": false }, { "ref": 22, "data_element": "Vehicle-Specific Data including VIN characteristics", "used": false }, { "ref": 23, "data_element": "Voice Analysis", "used": false }, { "ref": 24, "data_element": "Weather", "used": false }, { "ref": 25, "data_element": "Other: Non-Traditional Data Elements (Regulator to specify)", "used": true, "models": [ "claims-fastpath", "uw-triage" ], "how_used": "Decision input at inference (output use); training/testing use per the carrier's model documentation.", "internal_source": true, "third_party_vendors": [] } ] } ], "data_dictionary": "available on request from the carrier; RegBench records feature classes, never raw values" }, "regbench_disclosure": { "title": "Third-party AI system disclosure: RegBench (Exhibit C format)", "items": { "1_name_version": "RegBench assessment engine regbench-engine/0.4.0; rules 2026.10+dbe5ddf6; corpus 2026-10-30-5f07c521", "2_use_cases": "Regulatory assessment of AI-influenced insurance decisions; evidence assembly for examinations. Decision support only; the carrier remains the decision-maker.", "3_model_type": "Deterministic rules engine (rules as data). No machine-learning model produces findings. Optional narrative polish by a pinned language model is disabled in this configuration.", "3a_agentic": { "agentic": false }, "4_implementation_date": null, "5_development": { "vendor": "third_party", "vendor_name": "RegBench (Kenbright Actuarial and Financial Services North America Inc)" }, "6_purchased_modifiable": { "purchased": true, "developed_for_company": false, "modifiable": "doctrine-layer findings configurable as advisory; black-letter rules not modifiable by the carrier" }, "7_model_risks": "False negatives where a jurisdiction, line or stage is outside rule coverage (see coverage map); false positives on doctrine-layer findings (advisory).", "8_limitations": "Assesses recorded decision facts, not model internals; coverage limited to the states, lines and stages in the rule pack; provisions marked 'verification pending' are drafts.", "9_automate_augment_support": "support", "10_testing_validation": "Rule-level unit tests with positive and negative cases; citation integrity fails closed (100% existence check); point-in-time tests; eval set of showcase cases.", "10a_proxy_evaluation": "RegBench does not infer demographic characteristics. Portfolio tests consume carrier-produced proxy labels inside the carrier environment and release aggregates only.", "11_last_test_date": "2026-10-30", "12_financial_statement_effect": "None (no financial transactions generated).", "13_legal_compliance_review": "Rule pack reviewed against primary sources with verification status recorded per provision.", "14_regulatory_actions": "None." } }, "ledger": { "verification": { "ok": true, "entries": 690, "first_bad_seq": null, "head": "ecb444b6d2530cf62784043ce02a0e2591a68aa8108e2e685a473bef736e16d7", "reason": null }, "entries_in_chain": 690, "head_hash": "ecb444b6d2530cf62784043ce02a0e2591a68aa8108e2e685a473bef736e16d7" }, "totals": { "assessments": 577, "case_files": 547, "high_risk_models": 4 }, "_pack_hash": "c56fad6b39de203033bd60e843014f878c2cefc5e1ac9c22ef9f6c754297a454", "_ledger_seq": 1375, "_submitted_to": null }