AI Police Report Laws by State: 2026 Tracker

The short answer

As of June 2026, U.S. states regulate law-enforcement AI in four overlapping areas, and the map above lets you filter by each. For AI-written police reports, two states have enacted laws — California (SB 524) and Utah (SB 180) — both requiring disclosure and officer certification, with California also mandating that agencies retain the original AI draft and keep an audit trail; New York and Minnesota have bills pending, and prosecutors in Connecticut and King County, Washington have imposed restrictions. For police facial recognition, roughly 15 states have enacted limits (warrant requirements, “not the sole basis for arrest” rules, or outright moratoria), with more bills pending. For investigative AI, dedicated law is still emerging — New York’s broad “Covered AI” bill is the furthest along. And several public-sector AI-governance laws (Texas’s TRAIGA, California, New York’s LOADinG Act, Connecticut, Maryland, Vermont, Kentucky, Indiana) reach law-enforcement AI even when they don’t name it.

This tracker is a research compilation, updated June 2026. It is not legal advice — verify the current text and status of any law with the official source before relying on it. Laws here are changing quickly.

How to read the map

In the All laws view, a state is green only when every law it has is enacted. If a state has an enacted law and something pending or restricted in another category, it shows as Mixed — open that state to see each category broken out. Switch the filter to a single category (Police reports, Investigations, Facial recognition, or Gov’t AI) to see the precise status for just that category.

AI police report laws

California — SB 524 (enacted, effective Jan 1, 2026). The most comprehensive in the country. When a report is generated wholly or partly by AI, the agency’s policy must ensure disclosure of the AI program used, an officer’s certification that they reviewed it for accuracy, retention of the first AI-generated draft for as long as the final report, and an audit trail tied to the source footage or audio. Only the final report counts as the officer’s statement, and vendors are barred from reselling agency data.

Utah — SB 180 (enacted, effective May 7, 2025). The first state to act. Requires a written generative-AI policy, a disclaimer on any AI-assisted report, and officer certification of accuracy. Utah does not require retaining the draft.

Pending: New York’s S10425 / A9253-A is the most ambitious — its “Covered AI” definition expressly includes generative report-writing AI, with a public inventory, disclosure, first-draft retention, and both Attorney General and private enforcement. Minnesota’s SF 4575 / HF 4536 would prohibit generative AI in official records, including police reports.

Restricted by prosecutors (no statute): Connecticut’s Chief State’s Attorney imposed a statewide moratorium on using AI to draft or narrate criminal reports (April 2026). King County, Washington’s Prosecuting Attorney’s Office won’t accept AI-assisted report narratives (memo, Sept 2024). Virginia’s HB 1294 was carried over to 2027; Nevada’s SB 199 ban failed in 2025.

Police facial recognition laws

Roughly 15 states have enacted limits on law-enforcement facial recognition, and the approaches cluster into a few patterns. Warrant or court-order requirements appear in Washington (RCW 43.386), Montana, Massachusetts, Maryland, and Utah. “Not the sole basis for arrest” safeguards appear in Alabama, Maryland, Montana, and Washington. Near-total moratoria exist in Maine (broadest government ban, narrow carve-outs) and Vermont (only a child-exploitation exception). Body-camera-only prohibitions — barring facial-recognition analysis of body-cam footage — are in place in Minnesota, Oregon, and New Hampshire. Kentucky makes evidence from non-compliant FR use inadmissible.

Bills are pending in Illinois, Wisconsin (where Milwaukee PD adopted an administrative ban), South Carolina, West Virginia, and Wyoming. Proposals failed in Florida, Georgia, Idaho, and Hawaii. And in states without statutes, courts have stepped in: New Jersey (State v. Arteaga) requires defendants be notified of FR use, and a Michigan settlement bars Detroit PD from arresting on a face match alone.

AI in investigations

Dedicated law for investigative AI — predictive policing, biometric matching, automated license-plate readers, and AI-assisted case analysis — is the least developed of the four areas. New York’s “Covered AI” bill (S10425 / A9253-A) is the furthest along, requiring a public inventory and disclosure of investigative AI systems. Most other constraints flow indirectly from facial-recognition statutes and the public-sector AI-governance laws below.

Government / public-sector AI laws

Several states regulate government AI broadly enough to reach law-enforcement use, even without naming police. Texas’s HB 149 (TRAIGA, effective Jan 1, 2026) covers government AI and bans social scoring and rights-infringing biometric identification — though a security/investigation carve-out exempts much law-enforcement use. New York’s LOADinG Act bars state agencies from rights-affecting automated decisions absent authorization, with impact assessments and human oversight. Connecticut (PA 23-16), Maryland (SB 818), Vermont (Act 132), and Kentucky (SB 4) require agencies to inventory and assess high-risk and generative AI, and Kentucky’s registry captures the State Police. California (AB 302 / SB 896) and Washington (Executive Order 24-01) add inventory and disclosure duties and flag law enforcement as high-risk. Indiana (SB 150) created an AI task force and agency inventory.

One trap worth flagging: Colorado’s AI Act (SB 24-205) was repealed and replaced by SB 26-189 (effective 2027), which applies only to the private sector and does not reach law enforcement. Colorado does, however, have an enacted facial-recognition law (SB 22-113).

State-by-state at a glance

Status of enacted laws, pending bills, prosecutor restrictions, and failed proposals as of June 2026. Investigative AI is covered in the section above (only New York has a dedicated bill). “—” means no specific law or bill in that category.

* Colorado’s AI Act (SB 24-205) was repealed and replaced by SB 26-189 (effective 2027), which is private-sector only and does not reach law enforcement. Colorado’s enacted facial-recognition law (SB 22-113) still applies. States not listed (e.g., Alaska, Oklahoma) have no specific law or bill, though some agencies have set local policies.

What this means for agencies

The direction of travel is consistent across all four areas: disclose when AI is used, keep a human accountable through review and certification, preserve the original AI output and an audit trail, limit biometric identification, and don’t let a vendor own or resell agency data. Agencies evaluating any AI tool — for reports, investigations, or analysis — should ask whether it retains every draft, produces an auditable record tied to the source evidence, limits biometric matching appropriately, and can run in a CJIS-compliant environment. Those are precisely the requirements states are writing into law.

A note on our perspective: Barricade AI builds multimodal, auditable AI for law enforcement, so we have a commercial interest here. We keep this tracker factual and sourced because an honest accounting of the law serves agencies better than spin — and because the standards these laws converge on are the standards any responsible system should meet.

Frequently asked questions

Which states have laws on AI-written police reports?

As of June 2026, two states have enacted laws: California (SB 524, effective January 1, 2026) and Utah (SB 180, effective May 2025). New York and Minnesota have pending bills, and Connecticut and King County, Washington have prosecutor-led restrictions.

What does California SB 524 require?

Disclosure of the AI program used, an officer’s certification that they reviewed the report, retention of the original AI-generated draft for as long as the report is kept, an audit trail, and a prohibition on vendors selling agency data.

How many states regulate police facial recognition?

Roughly 15 states have enacted limits as of June 2026 — including warrant requirements (Washington, Montana, Massachusetts), “not the sole basis for arrest” rules (Alabama, Maryland), and near-total moratoria (Maine, Vermont). Several more states have bills pending.

Are AI-written police reports banned anywhere?

Not by statute, but Connecticut’s Chief State’s Attorney imposed a moratorium, and King County (WA) won’t accept AI-assisted report narratives. A statutory ban in Nevada (SB 199) failed in 2025.

Do broad state AI laws apply to law enforcement?

Sometimes. Public-sector AI-governance laws in Texas (TRAIGA), New York (LOADinG Act), Connecticut, Maryland, Vermont, Kentucky, and Indiana can reach law-enforcement AI — though some, like Texas, carve out security and investigative uses. Colorado’s revised AI Act (SB 26-189) is private-sector only and does not apply to police.

Is there a federal law on law-enforcement AI?

No. There is guidance (DOJ COPS Office, IACP) and a proposed Federal Rule of Evidence (FRE 707) on machine-generated evidence, but no binding federal law specific to law-enforcement AI as of June 2026.

Sources are linked per state in the interactive map above. Last updated June 2026. Not legal advice.