Many Pilot Companies Found the NAIC AI Supplement's Instructions Unclear

NAIC By InsureAI Wire. Published November 13, 2026.

Source: NAIC official document, content.naic.org (opens in a new tab)

Many companies that answered a survey on the NAIC’s pilot of the AI Risk Evaluation Supplement found its instructions unclear. The line comes from the “AI Pilot Summary” in the slides for the November 16 session of the Big Data and Artificial Intelligence (H) Working Group. The slides carry an update date of November 12 on the working group’s page, and Coral Manning of Wisconsin is scheduled to present the summary at the session that also considers adopting version 7.0.

The summary names 12 pilot states: California, Colorado, Connecticut, Florida, Iowa, Louisiana, Maryland, Pennsylvania, Rhode Island, Vermont, Virginia and Wisconsin. A note on its timeline slide says each state’s timing would vary and that not every pilot may have been completed by September 30. Thirty or more companies took part. The survey was offered to all of them and asked about their experience with each exhibit, and 15 or more sent a response. Minutes from the group’s October 8 meeting, which we covered, had said a longer pilot update was planned for the Fall National Meeting, where the November 16 session takes place.

On the positive side, most companies reported that they could respond to inquiries using documentation they already had. Companies also reported that the supplement led to productive discussions with regulators in which they could tell their story. Besides unclear instructions, the critical lines say companies would like more flexibility in how to respond, such as providing company documents in lieu of completing the exhibits. Another reads “Coordination in use of Supplement could be improved,” without saying who would coordinate with whom. No count sits behind any of these lines, so “most” and “many” are as precise as the record gets.

Companies were split on whether they preferred to receive all the exhibits at once or in sequence. Their practices also varied on whether generalized linear models (GLMs) fall within the scope of their own AI governance frameworks. Related topics came up in oral comments at the working group’s November 2 meeting, recorded in draft minutes that are item 1 for adoption on November 16. ACLI spoke to how the exhibits should be used, sequentially and in a risk-focused way; APCIA said the supplement itself should exclude traditional GLMs; and NAMIC said GLMs should leave its scope.

In the same minutes, the working group’s chair, Nathan Houdek, a Wisconsin commissioner, said the version adopted at the Fall National Meeting “will be available for voluntary state use next year.” No further substantive changes will be made before adoption, he said. Discussion of placing it in the Market Regulation Handbook or the Financial Condition Examiners Handbook will continue next year, leaving that question where the October 8 minutes left it. Miguel Romero of NAIC staff said the NAIC is rolling out an eight-part webinar series for regulators on AI, the NAIC Model Bulletin and the supplement. The series ends in early 2027, and he said it is meant to promote consistent implementation. On our reading, if a version is adopted on November 16, the supplement’s requests would reach an insurer in 2027 only from states that choose to use it.

Status as of November 12, 2026. The November 16 session has not taken place. Manning has not presented the summary, the November 2 minutes are still a draft awaiting adoption, and version 7.0 has not been adopted.

Share

Go deeper

Keep tracking

Information aggregation and analysis, not legal advice. See our disclaimer.