California opens an AI survey of insurers due December 30
As sent to subscribers on November 18, 2026.
The California Department of Insurance has started asking insurers, through a survey and a public workshop, how they use AI. In Dallas, the AI Risk Evaluation Supplement, which gathers the exhibits regulators would draw on to question insurers about AI, was listed for adoption by an NAIC working group on November 16 and by its parent committee on November 17. On the morning of November 17, when we checked, the NAIC had posted no record of the outcome, and the committee was set to meet later that day, from 12:15 to 1:30 p.m. Dallas time. The committee’s page says minutes are posted about 10 business days after a national meeting ends, so the formal record may take that long.
This week’s story: What California is asking insurers about AI
The department’s November 13 release, Commissioner Lara launches statewide effort to establish consumer safeguards for AI in insurance, sends its Responsible Use of Artificial Intelligence survey to “a representative sample”: admitted companies or groups with 2025 California premiums of $250 million or more, plus selected admitted InsurTechs, and selected insurers operating through managing general agents.
The survey covers how insurers deploy AI, how they govern it, what part vendors play and what AI has changed in their operations, with responses due December 30. Lara adds that insurers “remain accountable for the systems they use,” vendor tools included.
Nothing here is a proposed rule. The release poses its own question, whether the department is adopting AI regulations now, and answers “No.” The December 15 workshop comes before the survey deadline, and our November 15 report covers the categories and the workshop’s questions.
What to do this week
We suggest two steps to insurers doing business in California:
-
Look for the survey if your company or group could fall in the sample, and find out who holds it. Then compare your AI inventory and governance documents with the survey’s four subjects (deployment, governance, the vendors’ role, operational effects) and, well before December 30, mark any you could not support in writing.
-
Decide by December 12 whether to send written workshop comments. The invitation says those comments may be left out of any rulemaking record that follows and that the department is not required to respond; a view meant for that record has to be presented again in the public comment period, if a Notice of Proposed Action opens one.
On the Docket
-
November 20, 2026: At NCOIL’s annual meeting in Sanibel, Florida, Ladder Life’s Cara St. Martin presents AI use cases in life insurance to the life insurance committee, 3:45 to 5:00 p.m. On November 21, Verisk’s Joe Wodark presents AI developments in insurance to the financial services committee, 9:00 to 10:45 a.m., and item 6 of the NCOIL-NAIC dialogue that follows is an update on the work of the NAIC’s big data and AI working group. The schedule is dated November 10 and subject to change.
-
December 12, 2026: Written comments on California’s AI workshop are due, sent to CDIRegulations@insurance.ca.gov as the department’s invitation directs.
-
December 15, 2026: California’s workshop runs online from 10:00 a.m. PST to noon PST, or ends sooner once all attendees who wish to comment have spoken.
-
December 30, 2026: Responses to California’s AI survey are due.
This week in brief
Governance
The NAIC committee’s November 17 agenda names no version of the supplement, while the working group’s November 16 item names 7.0. Draft November 2 minutes in the working group’s November 16 slides record its chair saying the version adopted this fall will be open to voluntary use by states next year, and our reports on the two agendas and the pilot summary have the rest.
McDermott Will & Schulte, writing before the Dallas meetings in NAIC releases version 7.0 of its AI Risk Evaluation Supplement Ahead of Fall National Meeting, says industry reaction to the proposed changes “has generally been positive.”
The NAIC’s vendor framework, which covers the outside data and models behind property and casualty pricing and underwriting, is “now officially exposed” in its November 4 version, according to slides the Third-Party Data and Models (H) Working Group presented on November 16; they say adoption will be considered “in the spring next year.” The group’s page showed no comment deadline for it on November 17, and our November 17 report covers what changed since July.
New Jersey’s A5494, which would forbid an insurer to hand AI “the final decision” on denying a claim under an auto, homeowners or flood policy, was reported out of the Assembly Science, Innovation and Technology Committee on November 16 and referred to the Assembly Financial Institutions and Insurance Committee. New Jersey Globe’s N.J. lawmakers move to regulate AI in insurance, housing and the workplace reports one abstention and every other committee member in favor, and our September report reads the bill’s text.
The NAIC’s adoption map for its AI model bulletin, restamped November 13, now lists Texas among the states that adopted it, citing the state’s June 2026 bulletin B-0003-26. Our state tracker now follows the November 13 map.
The Elsewhere page gathers more reading from other outlets.
From the Guides
For pricing and filing teams. The vendor framework above concerns the outside data and models behind a rate; our November 12 guide is about the rate itself. It explains how a state reviewer judges a model-based rate under the state’s rating law, expecting each rating variable to connect to expected losses or expenses, and notes that NAIC guidance binds no insurer unless its state adopts it: How State Regulators Review an AI Pricing Model in a Rate Filing.
Reply and tell us which of the survey’s four subjects your team would find hardest to put on paper.
– The Editor, InsureAI Wire