No. Separate the model bulletin, state action and the developing examination supplement.

The NAIC’s December 2023 model bulletin addresses AI-supported insurance decisions, governance and information regulators may request. Its effect must be assessed through the relevant state’s laws, bulletin or supervisory approach. (content.naic.org)

The NAIC adoption map dated 8 October 2026 records recent actions including Mississippi Bulletin 2026-9 on 22 July 2026 and North Dakota Bulletin 2026-2 on 5 October 2026. The map itself cautions that it does not establish that every listed action contains all model elements. Do not turn the map into a claim of uniform nationwide requirements. (content.naic.org)

What is the latest verified examination-tool development?

The working group exposed AI Risk Evaluation Supplement version 6.0 on 16 October 2026, with comments due 30 October 2026. A further public discussion is scheduled for 2 November 2026. At this research cut-off, those are draft-development milestones, not a completed adoption or a carrier compliance deadline. (content.naic.org)

The earlier pilot summary describes a March–September 2026 pilot intended to test the tool with insurers across different lines. Buyers should use that history to understand the direction of supervision, not assume the pilot text and version 6.0 are identical. (content.naic.org)

What should an examination-ready production supplier provide?

The model bulletin addresses third-party diligence, appropriate audit and regulatory-cooperation arrangements, validation and information about data provenance and model operation. It also recognises that insurers may demonstrate compliance through alternative practices. (content.naic.org)

This desk recommends a compact evidence pack for each production use case:

1. Purpose and boundary: what the system does, and which decisions it does not make.
2. Input map: inspection sources, document types, personal information and known input limitations.
3. Output lineage: a way to connect material estimate items or narrative assertions to evidence.
4. Change history: model, prompt, carrier-rule and estimating-configuration changes relevant to the output.
5. Testing record: representative files, material errors, reviewer disagreements and correction outcomes.
6. Human authority: who can approve, reject, override or escalate.
7. Vendor chain: subcontractors, processing locations, retention and regulatory-response contacts.

The list is a proposed procurement artefact—not a representation that the current NAIC draft mandates these exact seven headings. It is suitable for an Evidence-to-Approval Control Audit and should be maintained alongside the carrier’s wider AI governance programme.