What happened
Version 5.0 of the NAIC AI Risk Evaluation Supplement (renamed from the AI Systems Evaluation Tool) was exposed on 31 August 2026 for a 30-day comment period that closed 29 September 2026, with a verbal-comment public hearing scheduled 8 October 2026 and a final version targeted for adoption at the NAIC Fall National Meeting in November 2026. Version 5.0, informed by a 12-state pilot, expands scope to machine learning and general language models, clarifies the AI 'system' vs 'model' distinction, tightens materiality, adds an explicit model-inventory request, and adds questions on oversight of third-party models (Exhibit B).
Why it matters
This draft is effectively the operational AI-governance examination tool US state regulators will use to scrutinise insurer AI systems, with requirements that flow down to AI vendors and third parties embedded in insurer products. Insurers will need documentation on every in-scope model, including third-party and embedded models, and must cite document and page numbers supporting each answer.
Action needed
Insurance carriers and AI vendors serving them should prepare evidence packs (model inventories, governance controls, third-party oversight) matching Exhibit A/B requirements and monitor the 8 October hearing and November adoption.