NAIC Plans Two Comment Rounds for Its AI Risk Evaluation Supplement

By the chair’s account, two comment windows stand between the NAIC’s AI Risk Evaluation Supplement and the adoption its working group is aiming for. Commissioner Nathan Houdek of Wisconsin, who chairs the Big Data and Artificial Intelligence (H) Working Group, told the group’s August 13 session in Columbus, Ohio, that the supplement will be revised using feedback received during the pilot and input from subject matter experts. An updated version, he said, will be ready for a public exposure comment period in early September. Two exposure periods are planned for this fall, and verbal feedback will be taken at the public meetings that follow each of them, with the stated goal of finalizing the supplement and adopting the final version at the Fall National Meeting.

That sequence sits in the working group’s final minutes of the August 13 session, carried as Attachment 1 for the group’s August 31 meeting and stamped “Draft Pending Adoption,” so it is the record as it stands before adoption. No motion on the supplement itself is recorded. The one vote the document carries was on the group’s July 22 minutes, and that motion passed unanimously. Houdek returned to the date at the end of the meeting, under other matters, reminding participants that an updated version would be exposed for a public comment period at the beginning of September. One month is named in the record. The second window is placed in the fall with no date attached to it.

The 12 states taking part in the pilot have been engaged since March and have met nearly weekly to discuss how each state is implementing it. Each of them has also gone back to its own pilot companies through a centralized survey of the experience, and the chair said written feedback already in hand will help inform revisions to the supplement. The pilot itself runs through September. On our reading of that calendar, a carrier that has been answering the supplement inside a pilot state has already put something into the drafting record, while a carrier outside those states gets its first turn in September.

A carrier reading the September text should not take it for the whole of what a department will ask. Pilot states ran the supplement two ways, inside a market conduct or financial examination already scheduled, or as a standalone questionnaire. The minutes then record, in their own voice rather than the chair’s, that each state can modify the questions or add its own. We take the September text as a common baseline rather than a ceiling, and an exam readiness file built to it alone can still encounter questions it has not seen.

The minutes stop short of the revision. No draft text is recorded as coming before the working group that day, and nothing in the session’s account says what the revision does. A summary of what changed, circulated before the exposure draft posts, is describing a document the public does not have. This record names the instrument only as the AI Risk Evaluation Supplement, a change we reported earlier this month. What the questionnaire asks today, still filed under the AI Systems Evaluation Tool name it carried before, is in our standing guide.

Name the person who reads the exposure version the week it appears, and name a second reader who checks it against how your AI governance program is documented today rather than against the version of the supplement your team already knows. Bookmark the working group’s own page on the NAIC site; the minutes point to no posting location, so that page is where we would look and not one the record names. If your company answered inside a pilot state, pull what you sent through that state’s survey and read the new draft next to your answers. Anything you raised that did not move is the first comment to file, and the one to file again in the second window if it still stands. Put both windows in the compliance calendar now.