Three letters ask the NAIC to hold back Exhibits C and D

“Exhibits C and D be removed or deferred until further piloting can be conducted” is what the Insured Retirement Institute recommends for half the exhibits in the NAIC’s draft AI Risk Evaluation Supplement. The working group drafting the supplement has posted that letter with the other comments on version 5.0, and it meets in public on October 8.

This week’s story: what the comment letters ask the NAIC to take out, and to put in

The American Council of Life Insurers (ACLI) says “many pilot participants tested only Exhibit A” (p.8 of the 206-page comment file, whose page numbers we use throughout). It asks that Exhibit C, on high-risk models, and Exhibit D, on model data, “not be used broadly” until tested further. NAMIC, the mutual insurers’ association, “continues to suggest the removal of Exhibits C and D” (p.151). All three also ask that generalized linear models (GLMs) be taken out of the supplement (pp.5, 95, 150).

Other submissions ask for additions. Five NAIC consumer representatives write that “consumers would benefit from a uniform definition of high-risk models,” since Exhibit C leaves the criteria to each company (p.80). Connecticut’s page, headed “From Wanchin Chou (CT DOI),” proposes an Exhibit C item, 11a, on whether a company evaluates variables or outputs “for potential proxy relationships” (p.83).

AHIP, a national association whose members provide health care coverage, dated its letter October 2, three days after comments were due. Its redline, a marked-up copy of the draft, carries 16 margin comments; one calls the term “Direct” “not well-defined” and recommends “material” instead (pp.16, 17).

The file contains nothing from the working group in reply, and no text of a version 6.0. On October 8, listen for three answers. Do Exhibits C and D stay in the next draft? Do GLMs stay in scope? Is either proposed addition taken up? The submissions above, apart from AHIP’s, are quoted at length in our October 5 report on the comment file.

What to do this week

• Ahead of the working group’s October 8 call, mark which parts of your supplement preparation file address Exhibits C and D. Those are the two that ACLI, the retirement institute and NAMIC ask the NAIC to hold back or remove. Nothing is decided, so the work stays, and the marking shows what a revised draft could change for you.

• During that session, note anything said about when a next draft will be exposed and for how long. The working group’s August 31 timeline slide had a version 6.0 going out for 14 days of comment in October. That was a plan on a slide, and the group’s page listed no such draft when we read it on October 6. Settle now who would write a comment in a window that short and who would sign it.

• If your company sets its own criteria for which models count as high-risk, as Exhibit C in the draft leaves it to do, put them on one page with the date they were approved. Connecticut’s proposed 11a would ask for “the methodology used and any resulting action taken” on proxy checks, so confirm both are recorded wherever you run such checks. These are proposals in two submissions; neither is text of the supplement.

On the Docket

• October 8, 2026: The NAIC working group’s one-hour public Webex starts at 10:00 AM CT. Its page gives the purpose: “continue public discussions on the AI Risk Evaluation Supplement.” No agenda or materials were posted as of October 6. Whoever will take notes for you on the three questions in this week’s story. NAIC working group page

• October 9, 2026: The date McDermott Will & Schulte’s September 28 alert gives for comments on a draft AI bulletin from Arizona’s insurance department. We have not read a department notice ourselves. Carriers writing in Arizona, for which our state tracker has no AI bulletin on record. McDermott client alert

• October 26, 2026: Colorado’s Attorney General hears testimony on proposed rules under the Automated Decision-Making Technology and Chatbot Safety Acts in Denver and by video from 10:00 AM. Written comments close at 11:59 PM MT the same day. The rulemaking page says a revised draft will come “at least five days prior to the public hearing” and linked none as of October 6. Anyone waiting to comment on the revised text. Notice of hearing

• November 2, 2026: A second one-hour public Webex of the same NAIC working group, at 10:00 AM CT, listed with the same stated purpose as the October 8 session. The page does not say which draft it will take up. The page did not list this date when we read it on September 29. Whoever keeps your NAIC calendar. NAIC working group page

This week in brief

Governance

California has a rewritten Insurance Information and Privacy Protection Act, SB 354, approved by the Governor on September 27 and operative July 1, 2028, that counts “artificial intelligence systems that are capable of outputting personal information” among the formats personal information can take. Our October 6 report sets out the pricing-model purposes it lists and what it leaves unsaid about deleting from a model: California’s New Insurance Privacy Law Treats AI Systems as a Place Personal Information Lives. SB 354 chaptered text

From the Guides

Fifteen requests, twenty-five data categories. That is what Exhibits C and D ask in the version 5.0 draft, and our guide goes through what each one asks: The NAIC AI Risk Evaluation Supplement, Exhibit by Exhibit. It also has six cross-exhibit questions for testing whether your answers agree with each other, and a readiness worksheet for Exhibits A to D.

If a regulator sent your company the pilot version of these exhibits this year, which of the four did the request include? Tell us in a reply to the email this issue arrived in; we read every reply.

– The Editor, InsureAI Wire