What it is
If you write insurance business, or advise on it, in more than one state, you've probably noticed that the rules on AI in underwriting, pricing and claims don't live in one place. The NAIC adopted a Model Bulletin on the Use of Artificial Intelligence Systems by Insurers in December 2023, and states are taking it up one at a time — some issuing the bulletin close to as-is, some writing their own guidance, and Colorado going further with a binding statute and regulations rather than a bulletin.
AI Insurance Watch is Kcalbin's free tracker of that process. As of its most recent verification, it resolves 26 jurisdictions as having adopted the NAIC Model Bulletin, 4 — California, Colorado, New York and Texas — as having their own AI-in-insurance instrument, and 21 jurisdictions as having taken no insurance-specific AI action. Colorado is marked as the only one of those four with a binding statute plus regulation, rather than guidance.
Every row carries a source. Where the tracker opened the state's own bulletin or bill page, it's marked State DOI / primary. Where it could only confirm status through the NAIC's own adoption reference list rather than a state-hosted copy, it says so and names that as the basis. A state Kcalbin couldn't confirm from either source is shown as unverified rather than filled in with a guess. Each row also shows a last-verified date, because the underlying page can change after that date without it showing up yet.
Some sources are checked daily. Kcalbin's change monitor reads nine source URLs on a regular cycle. The tracker lists its jurisdictions and their status beyond that, but it does not claim to watch every listed jurisdiction for changes in real time — only the sources that monitor reaches.
What it does not do
AI Insurance Watch is not legal advice, and it doesn't try to be. It tells you what a state has published and when, with a link to go read it yourself — it does not interpret the bulletin, tell you whether your practices comply with it, or tell you what a given regulator will do with it in an exam or an enforcement action.
It also does not monitor every row daily. The automated change monitor behind the page reaches nine source URLs. For rows outside that nine, what you have is the last-verified date shown on the row — not a live feed, and not a stated schedule for how often each one gets rechecked. If a state issues an update after that date, the tracker will not reflect it until it is next checked.
It does not guess. Where Kcalbin could not confirm a jurisdiction's status from an official or NAIC-referenced source, the row says unverified rather than offering a plausible-sounding status in its place.
It does not forecast what a state will do next, and it does not tell you what any one state's bulletin means for your specific product, pricing model or claims process. A bulletin is guidance from a department of insurance; a statute like Colorado's SB21-169 is binding law with its own regulations. The tracker distinguishes between the two, but distinguishing them is not the same as telling you which one applies to you, or how.
In short
AI Insurance Watch exists because the rulebook on AI in insurance changes state by state, on no fixed schedule, and nobody sends a notice when it does. The tracker gives you a dated, sourced list of who has acted and how — bulletin, statute, or nothing yet — and a way to check it yourself rather than take Kcalbin's word for it. It does not read every state's regulatory page every day, and it does not tell you what to do about what it finds.