HL-UR-0914 HOLD
On 2026-09-11, ur-auto v2.0 (a third-party system from ClinPath AI (fictional)) produced 'adverse determination' in a TX health utilization review transaction. Inputs included clinical criteria, medical records. 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 5 rule(s) in force on 2026-09-11 and recorded 3 finding(s); the most serious is 'Adverse utilization-review determination without a physician's determination' (critical, L2).
HOLD: 3 finding(s) under rules 2026.10+c4982313, corpus 2026-10-30-9c5dad5a. Recommended hold for the carrier's compliance review before the outcome is communicated; the carrier decides.
Evidence (carrier asserted). Supplied by the carrier and not verified by RegBench: ai_system.governance_attested, ai_system.model_doc_ref, ai_system.testing_report_ref, inputs.data_source_refs.
Findings (3)
Adverse utilization-review determination without a physician's determination
Finding. An adverse determination on a request for health care services was produced by an automated decision system and no physician determination is recorded.
Basis. § 4201.156(a): "A utilization review agent may not use an automated decision system to make, wholly or partly, an adverse determination." Utilization review must be conducted under the direction of a licensed physician (§ 4201.152), and before an adverse determination the provider must have a reasonable opportunity to discuss the treatment plan with a physician (§ 4201.206). The 2025 amendment applies to utilization review for plans delivered, issued or renewed on or after 1 January 2026.
Required action. Hold; route to a physician reviewer; give the provider of record the opportunity to discuss the treatment plan before any adverse determination issues.
Excerpt: tx-ins-4201-156 · state statute (government edict) · in force from 2025-09-01
"(a) A utilization review agent may not use an automated decision system to make, wholly or partly, an adverse determination. (b) The commissioner may audit and inspect at any time a utilization review agent's use of an automated decision system for utilization review. (c) This section does not prohibit the use of an algorithm, artificial intelligence system, or automated decision system for administrative support or fraud-detection functions."
Excerpt: tx-ins-4201-152 · state statute (government edict) · in force from 2007-04-01
"A utilization review agent shall conduct utilization review under the direction of a physician licensed to practice medicine in this state. The physician may not hold a license to practice administrative medicine under Section 155.009, Occupations Code."
Excerpt: tx-ins-4201-206 · state statute (government edict) · in force from 2007-04-01
"(a) Subject to Subsection (b) and the notice requirements of Subchapter G, before an adverse determination is issued by a utilization review agent who questions the medical necessity, the appropriateness, or the experimental or investigational nature of a health care service, the agent shall provide the health care provider who ordered, requested, provided, or is to provide the service a reasonable opportunity to discuss with a physician licensed to practice medicine in this state the patient's treatment plan and the clinical basis for the agent's determination."
Excerpt: tx-ins-4201-002-1c · state statute (government edict) · in force from 2025-09-01
"(1-b) 'Artificial intelligence system' means any machine learning-based system that, for any explicit or implicit objective, infers from the inputs the system receives how to generate outputs, including content, decisions, predictions, and recommendations, that can influence physical or virtual environments. (1-c) 'Automated decision system' means an algorithm, including an algorithm incorporating an artificial intelligence system, that uses data-based analytics to make, suggest, or recommend certain determinations, decisions, judgments, or conclusions."
Consequential AI-supported decision taken without a person reviewing and agreeing
Finding. An AI system in a high-risk class produced an outcome adverse to the consumer, and no person is recorded as having reviewed and agreed with the decision before action was taken.
Basis. Commissioner's Bulletin B-0003-26 (12 June 2026) states that for consequential decisions "TDI expects a person to review and agree with all decisions before action is taken", that decisions must not be inaccurate, arbitrary, capricious or unfairly discriminatory, and that the expectation extends to third parties working with the regulated entity.
Required action. Route to the responsible reviewer before the outcome takes effect; record the reviewer on the transaction.
Excerpt: tx-tdi-b-0003-26-human-review · regulatory bulletin addressed to all regulated entities and their agents (government edict) · in force from 2026-06-12
"TDI expects a person to review and agree with all decisions before action is taken" on consequential decisions made or supported by AI; decisions must not be "inaccurate, arbitrary, capricious, or unfairly discriminatory"; "These expectations extend to any third party working with a regulated entity." The bulletin cites Insurance Code chapters 541, 542, 544, 560, 751, 401, 831, 4001, 4101 and 4201.
Excerpt: tx-tdi-b-0003-26-governance · regulatory bulletin addressed to all regulated entities and their agents (government edict) · in force from 2026-06-12
"All regulated entities using AI should be able to furnish procedures and protections upon request by TDI." TDI "will monitor the use of AI through examinations and product filings", will investigate consumer complaints about AI, and may inquire into governance frameworks, risk management, data and privacy protections, internal controls and specific uses of AI. Under Insurance Code chapter 4201, utilization review agents may not use AI to make an adverse determination.
Adverse outcome from a high-risk AI system with no human review
Finding. A high-risk system produced an adverse outcome that took effect without human review.
Basis. Doctrine: examiners applying the NAIC bulletin's governance and control expectations, and the AI Risk Evaluation Supplement's Exhibit B question on whether there is a human in the loop for direct-consumer-impact uses, ask first about human oversight of adverse automated outcomes; an unreviewed adverse decision is the case file most likely to be sampled.
Required action. Route to the carrier's reviewer; record the review and any override on the transaction.
Excerpt: naic-ai-bulletin-s3-1-ais-program · NAIC model guidance; binding in a state only as adopted (see adoptions register) · in force from 2023-12-04
"all Insurers authorized to do business in this state are expected to develop, implement, and maintain a written program (an 'AIS Program') for the responsible use of AI Systems that make, or support decisions related to regulated insurance practices. The AIS Program should be designed to mitigate the risk of Adverse Consumer Outcomes ... The Department also encourages the development and use of verification and testing methods to identify errors and bias in Predictive Models and AI Systems, as well as the potential for unfair discrimination in the decisions and outcomes resulting from the use of Predictive Models and AI Systems."
Point in time
- Law as of
- 2026-09-11
- Rules version
- 2026.10+c4982313
- Corpus snapshot
- 2026-10-30-9c5dad5a
- Engine
- regbench-engine/0.5.0 · deterministic-1
- Fingerprint
- 548497b88240674950bee063 (pattern grouping, not an integrity check)
- Input digest
- 6dea12327e60366d8baa7b4eb60a57e23bcf2434a4627931f69bd44723841fa6
- Evaluated
- 2026-10-31 05:42:55 UTC
- Exhibit C case file
- yes
Ledger
- Entry
- #6417 · assessment · recorded 2026-09-11T11:46:09+00:00
- Hash
- fdfd37fafe64b4c6164f3ee56aa08bbfbff3522d43d8fd3c0bc793815eec7db9
- Previous
- 7a7349d7aa91ddf28ba7d3ae4cb1e237c77a41e9a6ad6cd3c4276567025bdfeb
- Stored copy intact
- yes, hashes match
- Assessed input intact
- yes, revision 1 matches the input digest
Carrier review
Resolved by physician · outcome remediated · 18.7 hours after it opened · Resolved in the carrier's review queue (synthetic).
Recorded facts revision 1
As asserted by the carrier. References are recorded, not verified: RegBench holds no copy of the documents they name.
- AI system
- ur-auto v2.0 · third party (ClinPath AI (fictional)) · high-risk class: utilization review
- Governance attested
- True
- Testing report
- test://harborline/ur-auto/2026Q2
- Model documentation
- doc://harborline/ur-auto/2.0
- Feature classes
- clinical_criteria, medical_records
- Protected factors used
- none
- External data (ECDIS)
- False
- Consumer report / credit
- False / False
- Data sources
- ehr://harborline/clinical-record
- Decision
- adverse determination · score 0.779 · codes MN-02
- Human review
- none
- Adverse to consumer
- True
- Notice
- not recorded · reasons: True · specific: True · external data disclosed: False
- Appeal channel
- –
- Consumer token
- tok_0a021ff4d3