AI in home insurance: the rules on underwriting and claims

There is no federal law written for AI in home insurance. The rules come from existing state insurance law (unfair discrimination, rate filing, unfair claims practices, consumer reporting) plus a National Association of Insurance Commissioners model bulletin, adopted in December 2023, that tells insurers to govern their AI systems and stay accountable for vendors' models. The practical effect: an insurer can use a model, but the model's output does not excuse an unlawful decision, and you can still ask for the reason in plain terms.

Where AI actually shows up in a home policy

Most of what gets called AI in insurance is statistical modeling that has been around for decades, joined by newer machine learning and image analysis. A homeowner is likely to meet it in five places.

• Underwriting. Aerial and satellite images are used to flag roof condition, a pool or debris, and tree overhang. See aerial imagery: how insurers inspect roofs from the sky.
• Risk scoring. A wildfire risk score or a catastrophe model output feeds eligibility and price.
• Pricing. Rating plans combine dozens of variables, including an insurance score; the more data, the more chances for a proxy to creep in.
• Claims triage. Software sorts claims into fast-track, desk-adjuster or field-inspection paths and flags possible fraud.
• Estimating. Photo-based tools, drone imagery and automated estimates; see AI, drones and automation in home claims.

This guide is about the rules. How the technology works day to day is in the linked guides.

The NAIC model bulletin

The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on December 4, 2023. A bulletin is guidance that tells insurers how the regulator will read existing law, and it binds only where a state adopts it. The NAIC's adoption map, current as of October 8, 2026, lists 27 states plus the District of Columbia and Puerto Rico as having adopted the bulletin. The most recent were Mississippi (July 2026), South Carolina (September 2026) and North Dakota (October 2026). The map lists four more states, California, Colorado, New York and Texas, as having their own insurance-specific AI regulation or guidance instead. Check the map for your state, because adoptions keep coming.

In substance, the bulletin says:

• An insurer remains responsible for compliance with unfair trade practice and unfair claims settlement laws, and its decisions must not be inaccurate, arbitrary, capricious or unfairly discriminatory, whether a person or a model makes them.
• The insurer should have a written governance program for its AI systems, proportionate to how it uses them, covering risk management, testing and documentation.
• Insurers that buy third-party data or models must do due diligence and should be able to audit those vendors by contract. The bulletin does not regulate vendors directly; the NAIC set up a Third-Party Data and Models task force to develop a framework.
• Examiners can ask for the documentation during an investigation or market conduct exam.

A bulletin does not create a private right to sue. Its force comes through the regulator.

State rules that go further

A handful of states have gone beyond the bulletin.

Colorado. Senate Bill 21-169, enacted in 2021, bars insurers from using external consumer data, algorithms or predictive models in ways that unfairly discriminate on protected characteristics, and it directs the commissioner to adopt rules line by line. The rules we could confirm at this writing apply to life, private passenger auto and health benefit plan insurers. We did not find a homeowners rule under that law, so check the Colorado Division of Insurance for the current status.

California. Prop 103 already makes rating factors subject to prior approval by the insurance commissioner, and the state's catastrophe-model and wildfire rules require insurers to disclose wildfire risk scores and let homeowners appeal them. A rate filing based on a model has to be justified to the department. See California's Sustainable Insurance Strategy.

Other states. New York issued Insurance Circular Letter No. 7 on AI and external data in July 2024, and Texas issued Bulletin B-0036-20 in 2020, according to the NAIC's adoption map. Several legislatures also consider AI bills each year. Search your state on the home insurance rules by state and see home insurance legislation to watch.

Because most states apply their existing law, the first question is usually whether the practice violates a long-standing rule, not whether a new AI statute exists.

Existing laws that already apply

Four older rules matter more to a homeowner than any new AI statute.

1. Rate laws. A rate has to be filed, and in prior approval states approved, and cannot be excessive, inadequate or unfairly discriminatory. A model that changes a rating factor generally changes a filing. See how insurers file rate increases with regulators.
2. Unfair discrimination and fair housing. A variable that is a proxy for race or another protected class can expose an insurer even if no one meant it to. See insurance redlining and fair housing law.
3. The Fair Credit Reporting Act. If an insurer takes an adverse action, such as denying coverage or charging more, based in whole or in part on a consumer report like a credit-based score or a CLUE report, it must tell you, name the reporting agency, and explain your right to a free copy and to dispute errors. See insurance scores: how credit affects home insurance and CLUE reports: how claims history follows you and your house.
4. Claims handling rules. The deadlines and duties in unfair claims settlement practices laws apply whether an adjuster or a program makes the call. An automated denial letter still has to cite the policy provision.

What the debate looks like

The case for AI is that better data improves accuracy: faster claim payment on small losses, fewer needless inspections, and prices that track risk more closely, which can lower rates for lower-risk homes. Insurers also say automated triage frees adjusters for complex claims.

The case against is that opaque models can embed errors and bias, can be hard for a regulator or consumer to test, and can be wrong about your roof with no easy path to correct it. Consumer advocates also warn that automation can speed up denials and lowball offers. Neither side has the full evidence: studies of outcomes by company are limited, and insurers treat model details as trade secrets. Regulators are the ones who can see the files.

If a model decided something about your home

1. Ask the insurer for the specific reason in writing: "roof condition per aerial image," "wildfire score of X," "claims history."
2. Ask what data was used and how to correct it. If an image is stale, send dated photos, a roofer's letter or a repair invoice.
3. If a consumer report was used, request the adverse action notice and a free copy of the report.
4. For a claim, ask for the estimate line items, the photos used and the policy provision behind any denial. Ask for a human adjuster review in writing.
5. Escalate with the state insurance department if the answer is evasive. The process is in how to file a complaint with your state insurance department.

Keep your own record: dated photos of the roof and exterior, maintenance receipts and inspection reports are the best rebuttal to an automated flag.

Key takeaways

• There is no federal AI law for home insurance; state insurance laws and a 2023 NAIC model bulletin, adopted by 27 states plus D.C. and Puerto Rico as of October 2026, set the expectations.
• An insurer is accountable for its models and its vendors' models under unfair discrimination and claims practices laws.
• Colorado has the most developed state AI rules, but the ones we could confirm do not cover homeowners yet.
• You can ask the insurer for the reason, the data and a human review, and you can complain to the state department.
• Dated photos and repair records are the best answer to an automated flag about your roof or lot.