Where we stand. NowAccident exists for people who were in a crash. Everything on this page is free and sourced to court opinions, state regulator documents, and published research.
Quick answer
Regulators did eventually respond to AI in insurance. In December 2023 the National Association of Insurance Commissioners adopted a Model Bulletin on the Use of Artificial Intelligence Systems by Insurers.
As of the NAIC's own adoption map dated August 6, 2026, 25 jurisdictions have adopted it — 24 states plus the District of Columbia. Four more (California, Colorado, New York, Texas) run their own insurance-specific AI guidance instead.
That leaves 22 states with neither.
And here is the part that matters most to you, which almost no coverage mentions: the laws that actually require a human to review an AI-driven claim denial are, so far, health insurance laws. Auto injury claimants are largely outside them.
The thesis of this series in one line: the process that prices your injury was designed by consultants, tuned on past settlements, and built around whether you have anyone advising you.
Parts 1 through 4 were about a system built in the 1990s. This part is about what happened when that system got a new name.
The short version: regulators moved faster this time. They did not move toward you.
WHAT THEY DID
What is the NAIC AI model bulletin?
The NAIC is the organization of state insurance commissioners. It does not regulate anyone directly — it writes models that states choose to adopt.
On December 4, 2023, it adopted the Model Bulletin on the Use of Artificial Intelligence Systems by Insurers. In substance it tells insurers that decisions made or supported by AI must still comply with existing insurance law, sets expectations for how insurers govern AI internally, and tells them what a regulator may ask for during an examination.
Read that description again and notice what shape it is. It is a set of expectations about the insurer's internal program. It is not a statute. It creates no right that you, as a claimant, can exercise.
And it has no force at all until a state adopts it.
Which states have adopted it?
As of the NAIC's map dated August 6, 2026, these 25 jurisdictions have adopted the model bulletin:
Alaska · Arkansas · Connecticut · Delaware · District of Columbia · Hawaii · Illinois · Iowa · Kentucky · Maryland · Massachusetts · Michigan · Nebraska · Nevada · New Hampshire · New Jersey · North Carolina · Oklahoma · Pennsylvania · Rhode Island · Vermont · Virginia · Washington · West Virginia · Wisconsin
Four more jurisdictions have their own insurance-specific AI regulation or guidance rather than the model bulletin:
| State | Instrument | Date |
|---|---|---|
| California | Bulletin 2022-5 | June 30, 2022 |
| Colorado | 3 CCR 702-10 | Effective November 13, 2023 (amended, effective October 15, 2025) |
| New York | Insurance Circular Letter No. 7 | July 11, 2024 |
| Texas | Bulletin B-0036-20 | September 30, 2020 |
Twenty-nine jurisdictions with something. Twenty-two with nothing.
If you live in one of those 22, your insurance regulator has not published a framework specific to how insurers use AI. That is not a scandal — it is a fact about where you happen to live, and it is exactly why we are building a full state-by-state map as a separate project.
Why isn't California on the adoption list?
Because California went first.
California's Bulletin 2022-5 was issued on June 30, 2022 — a year and a half before the NAIC adopted its model. Commissioner Ricardo Lara's bulletin reminded insurers and licensees of their obligation to market insurance, charge premiums, investigate suspected false claims, and pay claims in a way that "treats all similarly-situated persons alike."
If you have been reading this series in order, one line in that bulletin should stop you cold.
The thing California's regulator said it was looking at
The 2022 bulletin describes examples of alleged unfair discrimination the Department was investigating. Among them:
- Claims from certain inner-city ZIP Codes being subjected to special scrutiny
- Facial recognition being used in claims decisions
- Collection of personal information unrelated to the risk being underwritten
The bulletin also notes that greater industry use of AI and algorithms had produced an increase in consumer complaints about unfair practices, and it flags concern about seemingly neutral characteristics functioning as proxies for prohibited ones.
Now go back to Part 3.
The 2010 regulators recorded that the claims valuation software ran across 119 tuning regions, with states split into separate regions wherever settlement values varied by more than 10%. Geography was an input into what your injury was worth.
In 2022, California's insurance regulator wrote that it was investigating claims from certain ZIP Codes being treated differently.
Thirty years apart. Same variable. We are not asserting these are the same practice or the same companies — we have no evidence for that and will not imply it. What we are saying is narrower and, we think, more useful: the variable that regulators are worried about in the AI era is one that was already documented as an input in the era before it.
Does a state adopting the bulletin give me any rights?
Here is where most coverage of this topic quietly misleads people, so we are going to be blunt.
| What the bulletin does | What it does not do |
|---|---|
| Sets expectations for insurers' internal AI governance programs | Give you a right to be told AI was used on your claim |
| Tells insurers what regulators may ask for in an examination | Create a right to human review of an AI-influenced decision |
| Reminds insurers that AI-supported decisions must follow existing law | Create any new penalty or private remedy |
| Applies where a state has adopted it | Apply at all in the 22 states that have not |
A bulletin is a message from a regulator to the companies it regulates. You are not a party to it. If you write to your insurer citing the NAIC model bulletin and asking whether AI touched your claim, nothing obliges them to answer.
That is the same wall Part 4 described, wearing newer clothes.
Do any laws require a human to review an AI claim denial?
If you have seen headlines about states banning AI claim denials, they were probably accurate — and probably not about you.
The wave of laws requiring a human being to review a denial has landed almost entirely in health insurance.
- Arizona HB 2175, signed in 2025 and effective July 2026, requires a physician licensed in the state to conduct an individual review using independent medical judgment before certain claims are denied. Health coverage.
- Louisiana SB 246, in the 2026 regular session, addresses health insurance issuers using AI or automated decision systems, including notice, human determination, and appeals. Health coverage.
- California's own AI-in-utilization-review legislation likewise concerns health insurers and licensed physician supervision. Health coverage.
In auto insurance, the picture is different. Florida legislators considered bills in the 2025 session to prevent insurers from using AI as the sole basis for a claim denial. They did not pass. The topic returned to a legislative subcommittee, where an industry representative made the argument you will hear repeatedly: if a practice is prohibited for a person, it is prohibited for AI.
That argument is not unreasonable. It is also not a disclosure requirement, a human-review requirement, or anything you can invoke.
So the honest map of your protections looks like this: if an algorithm affects your health claim, a growing number of states now say a person must look at it. If an algorithm affects your auto injury claim, mostly nobody has said that yet.
One more thing that did not happen
Colorado passed what was widely described as the most comprehensive state AI law in the country, SB 24-205, covering high-risk AI systems in consequential decisions including insurance.
Its implementation was delayed after industry pushback, and then, on May 14, 2026, it was repealed and replaced by SB 26-189, a narrower automated-decision-technology statute effective January 1, 2027.
The most ambitious state AI law in America never took effect. We are including that because a series that only reported the wins would not be worth reading.
What we could not verify
You may have seen a figure claiming that a very high percentage of auto insurers — often stated as 88% — use or plan to use AI.
We could not trace that number to a primary source we are willing to stand behind, so we are not publishing it. If you have seen it cited with a verifiable original source, tell us and we will correct this page.
What we can say without qualification: the NAIC's Big Data and Artificial Intelligence Working Group spent 2025 and 2026 developing an AI Systems Evaluation Tool for regulators to use during market conduct and financial examinations, precisely because regulators needed a structured way to ask insurers what their AI is doing.
Regulators do not build examination tools for a technology nobody is using.
WHAT IT MEANS FOR YOU
Your rights depend on your ZIP code. Twice.
This is the sentence that ties the whole series together, and it is literally true in two separate ways.
First: per the 2010 regulatory findings, the valuation software was tuned by geographic region — 119 of them — so where you live affected what your injury was calculated to be worth.
Second: as of August 2026, whether your state's insurance regulator has published any AI framework at all depends entirely on which state you live in. Twenty-nine have something. Twenty-two do not.
Same injury. Same records. Different rules, different scrutiny, different published data — because of a line on a map.
Were you in a crash? Start a free accident record at nowaccident.com. No signup, no email, nothing sold. Five parts, three eras, and the same finding each time: the rules got written about how companies should behave, not about what you are entitled to know. Your own record is the one thing nobody else controls.
The pattern across all five parts
Look at what happened each time.
| Era | What the record shows | What claimants got |
|---|---|---|
| 1990s | Claims redesigned by consultants; documents show a two-track approach and a strategy of waiting out those who resisted | Nothing. The documents surfaced through lawsuits a decade later |
| 2010 | First multi-state exam of claims software: 2 million claims of data reviewed, zero individual files opened | A disclosure letter, from one company, for about five years |
| 2026 | 25 jurisdictions adopt AI governance expectations; regulators build examination tools | Governance rules for insurers. No consumer right to ask |
Each round produced more regulation of the industry and roughly the same amount of leverage for the individual person with the injury: almost none.
That is not a conspiracy. It is what happens when the rules get written about how companies should behave rather than about what claimants are entitled to know. Nobody chose to leave you out. Nobody was arguing for you in the room.
So what actually gives you leverage?
Given all that, be clear-eyed about what does and does not help you.
Does not help you:
- Asking whether AI was used. Nothing compels an answer.
- Citing the NAIC model bulletin. You are not a party to it.
- Waiting for a law to arrive. In auto, it has not.
Does help you, and every one of these came from an earlier part of this series:
- A contemporaneous claim log (Part 1) — dates and specifics beat memory
- Complete, accurate medical records (Part 2) — the system cannot price what is not written down
- Your state's published complaint record (Part 3) — the regulator's own scoreboard
- Written requests with deadlines attached (Part 4) — in California, 10 CCR § 2695.7 has real numbers in it
- Knowing your filing deadline — the only irreversible mistake in the whole process
Nothing in the AI regulation of 2026 replaces any of those. It does not even come close.
WHAT YOU CAN DO TODAY
Find out what your state has actually adopted. Ten minutes.
This is the last "one thing" in the series, and it is the one that generalizes beyond California.
Do this:
- Open the NAIC's own adoption map — Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers. It is a two-page PDF, free, and lists every adopting state with its bulletin number and adoption date.
- Find your state. Three possible outcomes: it adopted the model bulletin, it has its own instrument, or it has neither.
- Write the result in your claim log with today's date — including "none" if that is the answer.
- If your state has something, note the bulletin number. That is the document your regulator uses when examining insurers in your state.
- Then check the date on the map itself. If it is much older than today, search your state's insurance department site directly. This list changes.
What to do with the answer — and what not to:
- Do not write to your insurer demanding compliance with a bulletin. You are not a party to it, and it will not produce anything.
- Do know which regulator has taken an interest in this, because that is the office that reviews claims-handling complaints in your state.
- Do understand that "my state has adopted nothing" is a real answer with real consequences, and it is worth knowing rather than assuming otherwise.
Why this one: every previous action in this series built your own file. This one tells you what exists outside it. After five parts, you should know exactly how much is there — and be able to plan around the fact that it is less than you were probably imagining.
Find out where you are in the process
Our free tool takes your situation — just crashed, offer received, denied, or stalled — and shows what typically happens at that stage, what you are likely to hear, and links you to your insurer's record in California's official complaint study.
No signup. No email. Nothing sold. Ever.
Was this useful? Then send it to someone who was in a crash. This site carries no advertising and sells nothing. It reaches people only when someone passes it on. Start a free accident record at nowaccident.com — no signup, no email.
The end of the series, and what we are doing next
Five parts. One consistent finding.
Every time this problem has been examined, the examination produced rules about how companies should behave — and almost nothing you can personally invoke.
The gap we keep hitting is the same one: nobody has systematically mapped what an ordinary claimant is actually allowed to find out, state by state. So that is what we are building next — all 51 jurisdictions, three questions each: does your regulator publish insurer complaint data, has it adopted an AI framework, and can you ask whether an automated system touched your claim.
We will publish the whole dataset, including the blanks.
Frequently asked questions
How many states have adopted the NAIC AI model bulletin? As of the NAIC's adoption map dated August 6, 2026, 25 jurisdictions have adopted it: 24 states plus the District of Columbia. Four additional jurisdictions — California, Colorado, New York, and Texas — have their own insurance-specific AI regulation or guidance instead. That leaves 22 states with neither. The NAIC updates this map; check it before relying on any count.
Has California adopted the NAIC AI model bulletin? No. California issued its own Bulletin 2022-5 on June 30, 2022, about eighteen months before the NAIC adopted its model bulletin in December 2023. The California bulletin addresses bias and unfair discrimination in marketing, rating, underwriting, and claims practices, including in the use of artificial intelligence and large data sets.
Does my insurer have to tell me if AI was used to decide my auto claim? Not under the NAIC model bulletin, which sets expectations for insurers' internal AI governance rather than creating consumer rights. We are not aware of any general requirement that an auto insurer disclose AI involvement in an individual bodily injury claim. If your state has enacted one, tell us and we will correct this page.
Is it legal for an insurer to use AI to evaluate my claim? Yes. The regulatory position across states has generally been that AI-supported decisions must comply with the same insurance laws that apply to decisions made by people — not that AI use is itself prohibited.
Do any states require a human to review an AI-driven claim denial? Several states have enacted or advanced such requirements, but they have applied predominantly to health insurance. Arizona's HB 2175, signed in 2025 and effective July 2026, requires individual review by a licensed physician using independent medical judgment for certain denials. Louisiana's SB 246 in the 2026 session addresses health insurance issuers using AI, including human determination and appeals. Bills applying a similar concept to auto insurers were considered in Florida's 2025 session and did not pass.
What is a model bulletin, and is it law? A model bulletin is guidance drafted by the NAIC that has no legal effect until an individual state's insurance department issues it. Once issued, it communicates a regulator's expectations to the companies it regulates. It is not a statute and generally does not create private rights for consumers.
What happened to Colorado's AI law? Colorado's SB 24-205, described as the most comprehensive state AI governance law in the country, had its implementation delayed and was then repealed and replaced on May 14, 2026 by SB 26-189, a narrower automated-decision-technology statute effective January 1, 2027.
What is the NAIC AI Systems Evaluation Tool? A structured framework developed by the NAIC's Big Data and Artificial Intelligence (H) Working Group across 2025 and 2026, intended to help state regulators gather information about an insurer's use of AI during market conduct and financial examinations.
Where can I check whether my own state has adopted anything? The NAIC publishes a two-page adoption map listing every adopting state, its bulletin number, and its adoption date. It is free and updated periodically. Because adoptions occur continually, check the date printed on the map itself.
Sources
Every figure verified against primary sources on August 8, 2026.
- National Association of Insurance Commissioners, Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers — adoption map, status as of August 6, 2026. Primary source for the list of 25 adopting jurisdictions, their bulletin numbers and adoption dates, and the four jurisdictions with their own insurance-specific instruments. content.naic.org
- NAIC, Artificial Intelligence topic page and Big Data and Artificial Intelligence (H) Working Group 2026 charges — the Model Bulletin's December 4, 2023 adoption, its content and purpose, and the AI Systems Evaluation Tool. content.naic.org/insurance-topics/artificial-intelligence
- California Department of Insurance, Bulletin 2022-5, Allegations of Racial Bias and Unfair Discrimination in Marketing, Rating, Underwriting, and Claims Practices by the Insurance Industry, issued June 30, 2022 by Commissioner Ricardo Lara — the obligation to treat similarly situated persons alike; the described investigations including claims from certain inner-city ZIP Codes and facial recognition in claims decisions; the noted increase in consumer complaints. insurance.ca.gov (bulletin PDF) · CDI press release, June 30, 2022
- Multi-State Market Conduct Regulatory Agreement, executed August 27, 2010, hosted by the Oregon Division of Financial Regulation — the 119 tuning regions and the 10% settlement-variation threshold referenced above. dfr.oregon.gov
- Arizona HB 2175 (signed 2025, effective July 2026) — individual review by a physician licensed in the state, using independent medical judgment. Health coverage.
- Louisiana SB 246, 2026 Regular Session — requirements for health insurance issuers using artificial intelligence or automated decision systems, including notice, human determination, and appeals. legis.la.gov
- Florida House Insurance & Banking Subcommittee proceedings and reporting on 2025-session bills addressing AI as the sole basis for claim denial, which did not pass.
- Colorado SB 24-205 and SB 26-189 — the repeal and replacement signed May 14, 2026, with the replacement effective January 1, 2027.
Not used as sources: law firm marketing pages and settlement-estimate services. Not published: the widely circulated figure for the share of auto insurers using AI, which we could not trace to a verifiable primary source.
Reviewed by: Alex Song, Ph.D. — Researcher
Last updated: August 8, 2026 · Data verified: August 8, 2026
Read this first, and note the date. Every figure below was verified against primary sources on August 8, 2026. This is the fastest-moving area in the entire series — bulletins get adopted monthly, bills die and return, and one major state law was repealed before it ever took effect. Check the NAIC's own map before relying on any count, including ours. We link to it directly. Nothing here says any insurer is doing anything wrong.
This is general information, not legal advice. It does not create an attorney-client relationship. Laws vary by state and change constantly — this area more than most. Consult a licensed attorney in your state.
Think something here is wrong? Tell us and we will correct it: support(a)nowaccident.com