NAIC Model Bulletin Adoption Now Spans Most States

As of the NAIC’s April 2026 adoption map, 24 states and the District of Columbia have formally adopted the AI Model Bulletin originally adopted December 4, 2023. Four more (California, Colorado, New York, and Texas) regulate insurance AI under their own rules rather than the model. The map is the authoritative status source, but the differences between each state’s version are where the practical work lies.

The headline number matters less than the footnotes under it. Adoption is not a single on/off switch. Some states adopt the model verbatim. Others attach interpretive guidance, state-specific timing, or additional notice requirements. A carrier that treats the bulletin as a uniform national standard will miss the seams where its own state has added or softened language. The first task is to map which states your company writes in, which version of the bulletin each has adopted, and whether the state has issued any companion guidance or circular letter that changes how examiners will read it.

The second task is to stop reading adoption as exemption. Even in states that have not adopted the model, AI-assisted insurance decisions still sit under existing unfair-trade-practices, unfair-discrimination, and unfair-claims-settlement statutes. The bulletin does not create new authority in states that adopt it; it tells examiners where to look. In states that have not adopted it, the underlying authority is still there, but the carrier has fewer clues about what the examiner expects to see. Having less to read makes the job harder.

The four independent states matter because they are usually the largest markets and the most active examiners. California, Colorado, New York, and Texas each have their own frameworks, and a carrier operating in all four is running four parallel compliance programs. Colorado runs two layers, and the newer one mostly points back at the older: SB 21-169 is what obliges insurers to test external data and models for unfair discrimination, and SB 26-189 deems a carrier already subject to it compliant in the practice of insurance from January 1, 2027. New York’s Circular Letter 7 sets specific expectations for underwriting and pricing AI. California Bulletin 2022-5 reaches wider than underwriting alone, naming marketing, rating, claims handling, and fraud investigation as well. Texas is working from a 2020 bulletin that makes the carrier answerable for the accuracy of third-party data feeding rating, underwriting, and claims. These are not minor variations. They are separate regulatory tracks that happen to share a topic.

Three things follow from that. First, pull the adoption status state by state and compare it against your footprint. Second, for each state where you write, identify the exact mechanism: model bulletin adoption, state-specific guidance, or independent legislation. Third, remember what the adoption count leaves out. The four states running their own regimes are not in it, so the number tells you how many regulators borrowed the NAIC’s language, not how high your program has to reach. Where it has to reach across a multi-state footprint is worked through in our AI governance guide.

The same logic applies to distribution and M&A. A carrier acquiring a smaller writer in a new state inherits that state’s bulletin version and any companion rules. A managing general agency or third-party administrator using your paper may create AI exposure in a state your compliance team has not mapped. The adoption count is a snapshot; the state footprint is a moving target. Keeping the two aligned belongs on the AI governance committee agenda as a standing item, revisited whenever the footprint moves.

The check worth running is a list of every state your paper touches, including the ones you reach only through an MGA or TPA, matched against which bulletin version each has adopted. That list is usually longer than the one compliance maintains. Reconcile the two now, using the rule-by-rule breakdown to see which version asks for what.