AI Regulation Tracker  /  For Insurance Professionals

Federal vs State AI Fight Hits Your Insurance Premiums

Washington is trying to preempt the state rules that govern AI pricing and underwriting, and insurers now have to build governance for two possible outcomes at once.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 1, 2026  ·  Last updated July 1, 2026

As of July 1, 2026, the federal effort to preempt state AI rules is not law. It rests on Executive Order 14365, a DOJ AI Litigation Task Force that has filed no cases, and a bipartisan House discussion draft for a three-year preemption that has not passed. Meanwhile more than half the states, 24 plus DC, have adopted the NAIC AI model bulletin governing insurer use of AI in pricing and underwriting, so insurers should govern to the stricter state standard while modeling what preemption would and would not erase.

If you run compliance or actuarial work at a carrier, you are being asked to govern your AI models against two futures at the same time. In one, the states keep the authority they have spent two years building. In the other, Washington preempts a big slice of it. Neither has happened yet, and that uncertainty is the actual news for anyone whose models decide what a customer pays.

The trigger is a June 20, 2026 Forbes analysis by Dara-Abasi Ita, What The AI Fight Means For Your Insurance Premiums, which lays out how a federal-versus-state clash over AI regulation lands directly on insurance pricing and underwriting. Here is the honest state of play, section by section, and what a working professional should do about it.

What Washington Has Actually Done

On December 11, 2025, President Trump signed Executive Order 14365, titled "Ensuring a National Policy Framework for Artificial Intelligence." It directs the Justice Department to go to court against state AI laws the administration judges too heavy a burden on the technology. The order created an AI Litigation Task Force inside the DOJ and gave it until January 10 to stand up.

Read the next part slowly, because it is the part people skip. As of Forbes' June 20 report, the task force exists, the attorney general has announced it, and it has not filed a single case. Forbes reports the task force is still waiting on a Commerce Department list of target laws before it moves. An executive order cannot strike down a state law on its own, which is why the order reaches for other tools: the litigation task force, a Commerce review to name offending laws, and a threat to withhold roughly $21 billion of federal broadband money from states that keep enforcing rules Washington dislikes, a figure Forbes attributes to the funding leverage in play.

There is also a legislative track, and it is only a draft. In early June 2026 a bipartisan House discussion draft surfaced that would block states from regulating AI development for three years. It is a proposal, not a law. State lawmakers have already urged Congress to reject it. So when you hear "federal moratorium," the accurate translation as of July 1, 2026 is: an executive order that cannot preempt on its own, a task force that has sued no one, and a preemption bill that has not passed. The threat is loaded but not fired.

The Quiet State Build-Out

While Washington argued about AI in the abstract, insurance regulators built the plumbing. Insurance in the United States is policed state by state, coordinated through the National Association of Insurance Commissioners, a body of the 50 state regulators. In December 2023 the NAIC published its model bulletin on the use of AI systems by insurers, setting expectations on governance, testing, and oversight. A model bulletin is not law by itself; each state has to adopt it.

By the NAIC's own count from its spring meeting in March, as reported by Forbes, 24 states and the District of Columbia have adopted the bulletin, with four more writing their own AI-specific rules. That is more than half the country, arrived at without much noise. The common thread: an insurer using AI in decisions that affect customers must keep a written program, govern the models, watch for unfair discrimination, and answer to examiners when asked. These are not aspirations on a notice board. They are expectations an examiner can walk in and test. If you want the deeper version of that rulebook, we covered it in the [NAIC AI model bulletin briefing](/ai-regulation-news/naic-ai-model-bulletin-insurer-governance-2026).

Where The Algorithm Meets The Premium

The reason this reaches a premium sits in two words: underwriting and pricing. New York wrote the sharpest rule. In July 2024 its Department of Financial Services issued Insurance Circular Letter No. 7, covering insurers that use AI and external consumer data in the underwriting and pricing of insurance policies and annuity contracts. Annuities are named on purpose, because that is the product millions of people use to turn savings into retirement income.

The New York letter tells insurers they must test whether an AI model produces a disproportionate adverse effect on protected groups, and that they cannot hide behind a vendor's black box if it does. In plain terms, if a pricing model quietly charges some customers more for reasons that track a protected characteristic, the insurer is responsible for finding that out and fixing it. Colorado went broader, with an AI Act covering AI across insurance practices rather than pricing alone, and that law took effect on June 30, 2026. AI has reached the claims side and the reserving side too, so a model that flags a claim or helps set reserves shapes both whether you get paid and whether the carrier stays solvent enough to pay you.

The Tool That Turns Rules Into Inspections

Rules on paper are one thing; inspections are another. The NAIC's Big Data and Artificial Intelligence Working Group has built an AI Systems Evaluation Tool, a standard kit examiners use to look under the hood of an insurer's models. By the NAIC's count as reported by Forbes, it was being piloted by 12 states as of March, with wider adoption expected at the NAIC's autumn meeting. This is the unglamorous part where guidance turns into someone with the authority to ask hard questions and a checklist to ask them from. For carriers, it means model documentation you could defend on paper now has to survive a live examiner. Our [NAIC exam pilot briefing](/ai-regulation-news/naic-ai-exam-pilot-insurer-governance) walks through what those examiners are trained to probe.

Why This Is A Two-Outcome Problem

Here is the strategic bind. If the state rules hold, an insurer using AI to price an annuity or a motor policy has to test that model for unfair effects and answer for it. If the rules fall or freeze under federal pressure, that obligation thins out and the model goes back to being a box only the insurer can see inside. The same algorithm gets watched closely or barely at all depending on a fight playing out in courtrooms and a Commerce Department memo.

For a compliance leader, the mistake is betting the house on either result. Preemption has not passed. The task force has sued no one. And even a successful federal preemption of state AI development rules would not automatically wipe out insurance-specific consumer protection, unfair discrimination law, or the McCarran-Ferguson Act's presumption of state primacy in insurance regulation, all of which sit on older legal ground than the current fight. The safer read is that you build to the stricter standard and keep a memo on what a preemption would and would not touch.

How To Hedge Either Outcome

This is the usable part. A short, defensible playbook that holds up whichever way the federal fight breaks:

• Inventory every AI or external-data model that touches underwriting, pricing, claims triage, or reserving. If you cannot list them, you cannot govern them.
• Assign each model a named human owner who can explain, in writing, what it does and why its outputs are fair. New York's rule exists partly to force that explanation.
• Run and document disproportionate-impact testing on protected classes now, on your highest-stakes pricing models first. This is required under the strict state rules and is simply good practice under the loose ones.
• Confirm which of your operating states has adopted the NAIC bulletin or written its own rule, and tier your compliance to the strictest state you write in rather than the weakest.
• Keep vendor contracts that give you audit rights into third-party models. "The vendor's black box did it" is not a defense any regulator has accepted.
• Write a one-page preemption memo: which of your controls are driven by state AI rules, which by unfair-trade-practice and anti-discrimination law that preemption would likely not reach, and what you would keep running regardless.

Build it to the higher bar and a federal rollback costs you nothing you needed. Build it to the lower bar and a state examiner can hand you a finding. The asymmetry favors governance.

The Honest Bottom Line

As of mid-June 2026, Forbes reports that no state insurance AI rule has been challenged in court, let alone overturned. Anyone telling you the states have already won or already lost is ahead of the facts. The professional move is not to guess the outcome. It is to run AI governance that a state examiner would pass and that a federal preemption would leave mostly intact, then revisit the memo when the Commerce Department names its targets or the House draft moves. If you want the working knowledge behind that, our [AI course for finance and insurance professionals](/leveraged-cpa-finance) covers model governance, documentation, and defensible AI use, and our [two-minute quiz](/quiz) points you to the right starting track.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.