NAIC Advances AI Examination Framework for Insurers
The National Association of Insurance Commissioners updated its AI examination supplement on Aug. 13, moving AI governance guidance toward evidence requests used in insurer reviews, PYMNTS reports. A 12-state pilot is testing the structured questionnaire through September, with feedback intended to inform subsequent public exposure drafts. The tool is not a certification program or new insurance law, according to PYMNTS.
The National Association of Insurance Commissioners (NAIC) updated an examination supplement for insurer AI oversight on Aug. 13, as a 12-state pilot tests a common questionnaire for gathering evidence about AI use and governance, PYMNTS reports. The document is intended for market-conduct reviews, financial examinations, financial analysis, and stand-alone inquiries.
PYMNTS reports that the NAIC working group renamed the document, previously called the "AI Systems Evaluation Tool," to clarify that it is neither a certification program, rating system, nor new insurance law. Instead, the supplement operationalizes principles in the NAIC's 2023 AI model bulletin through requests for information that regulators can use during examinations.
Pilot and revision schedule
According to PYMNTS, the pilot includes California, Colorado, Connecticut, Florida, Iowa, Louisiana, Maryland, Pennsylvania, Rhode Island, Vermont, Virginia, and Wisconsin. Participating states have used different processes: some incorporated the supplement into scheduled examinations, while others used it as an ad hoc questionnaire.
The pilot is scheduled to continue through September, although PYMNTS reports the NAIC cautioned that not every participating state may complete work by Sept. 30. State feedback and initial insurer responses are slated to inform version 5.0, which is expected to enter a 30-day public exposure period in September. PYMNTS further reports that version 6.0 is scheduled for a 14-day exposure period before regulators consider version 7.0 at the fall national meeting.
Documentation becomes examinable
Waterstreet Company reports that, as of early 2026, 23 states and Washington, D.C. had adopted the NAIC's AI model bulletin. Its guide characterizes the pilot as an examiner questionnaire covering AI governance, risk management, and model oversight.
According to Waterstreet, the 2023 bulletin calls for a written AI Systems Program and identifies governance frameworks, risk controls, testing records, and adverse consumer outcomes among materials insurers could be asked to produce in investigations or market-conduct actions.
For data science and ML teams working in insurance, the immediate issue is less the existence of an AI model than the reproducibility of its governance record. Across regulated industries, examination-oriented frameworks commonly increase demand for model inventories, validation artifacts, data lineage, human-oversight procedures, third-party model documentation, and records of monitoring for harmful outcomes. The NAIC pilot provides an early view of how those controls may be requested in insurance supervision, rather than only described in internal policy documents.
Key Points
- 1The NAIC pilot turns high-level AI governance principles into structured evidence requests that insurers may encounter during examinations.
- 2Twelve states are testing the supplement through September, creating near-term input for public drafts and possible broader adoption.
- 3Comparable regulatory examinations typically make model documentation, validation records, data lineage, and outcome monitoring operational requirements rather than policy aspirations.
Scoring Rationale
The NAIC pilot is a notable development for insurance ML teams because it connects AI governance principles to examination evidence and documentation. Its direct scope is limited to insurance, but the framework addresses recurring practitioner concerns around model oversight, testing, and auditability.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
