Which US States Have a Binding AI-Specific Statute

The single most common error in writing about US AI regulation is describing a bill as though it were a law. This page is a dated snapshot as at 11 August 2026, scoped to the instruments covered in this part of the library, with each entry carrying the bill number, the codified citation where one exists, and the date it began or begins to apply.

How to read this snapshot

Three distinctions are collapsed constantly and each one changes what you owe.

• Enacted is not in force. Several statutes below were signed a year or more before their compliance dates. A signed statute with a future effective date creates a deadline, not an obligation.
• AI-specific is not the same as applies-to-AI. The statutes in the first table name AI or an automated tool. The ones in the second are general privacy or civil rights laws that bite on AI systems without mentioning them, and in most states those are the ones that will reach you first.
• Effective dates move. Colorado’s AI Act has already been delayed once by later legislation. A date in this table is the date as it stood when the page was updated; the legislature can change it, and has.

Enacted and AI-specific

• Illinois — Artificial Intelligence Video Interview Act. HB 2557, Public Act 101-0260, codified at 820 ILCS 42, effective 1 January 2020; amended by Public Act 102-0047 with effect from 1 January 2022 to add demographic reporting. Text at the Illinois General Assembly. Covered in the dedicated page.
• Maryland — facial recognition in interviews. HB 1202 (2020), codified at Md. Code Ann., Lab. & Empl. § 3-717, effective 1 October 2020. Bill page at the Maryland General Assembly.
• California — bot disclosure. SB 1001 (2018), codified at Cal. Bus. & Prof. Code §§ 17940–17943, operative 1 July 2019. Text at California Legislative Information.
• New York City — Local Law 144 of 2021. Not a state law; a municipal ordinance codified in the NYC Administrative Code, with enforcement by the Department of Consumer and Worker Protection beginning 5 July 2023. Guidance at nyc.gov.
• Utah — Artificial Intelligence Policy Act. SB 149 (2024), effective 1 May 2024, amended in the 2025 session to narrow the generative-AI disclosure duty and extend the Act’s repeal date. Bill history at the Utah Legislature.
• Colorado — Colorado AI Act. SB 24-205, signed 17 May 2024, codified at C.R.S. §§ 6-1-1701 to 6-1-1707. Its application date was moved from 1 February 2026 to 30 June 2026 by legislation passed in the August 2025 special session. Bill page at the Colorado General Assembly. See the effective-date page.
• Illinois — HB 3773. Amends the Illinois Human Rights Act, 775 ILCS 5, to address AI in employment decisions, with effect from 1 January 2026. See the dedicated page.
• California — AB 2013 and SB 942. AB 2013 requires generative AI developers to publish a training-data documentation summary from 1 January 2026. SB 942, the California AI Transparency Act, imposes provenance and detection-tool duties on large generative AI providers; it was amended in the 2025 session, which changed both its scope and its compliance date, so confirm the operative date on the bill page rather than from any summary. Both at California Legislative Information.
• California — SB 53. The Transparency in Frontier Artificial Intelligence Act, signed 29 September 2025, imposing framework-publication and incident-reporting duties on large frontier developers. See the dedicated page.
• Texas — TRAIGA. The Texas Responsible Artificial Intelligence Governance Act, HB 149, signed in June 2025 with effect from 1 January 2026. Bill page at Texas Legislature Online. See the dedicated page.

Beyond these, a substantial number of states have enacted narrow AI-adjacent statutes rather than general ones: election deepfake laws, non-consensual intimate imagery laws covering synthetic images, laws on digital replicas of performers, and statutes governing state government’s own use of AI. Those are counted in some published tallies of “states with AI laws” and not in others, which is the main reason two credible trackers can report very different numbers for the same date.

Privacy statutes that reach automated decisions

None of these mention AI. All of them constrain it, through a profiling opt-out, a data protection assessment duty, or both.

• Virginia — Consumer Data Protection Act, Va. Code § 59.1-575 et seq., in force since 1 January 2023. Profiling opt-out detail.
• Oregon — Consumer Privacy Act, ORS 646A.570–646A.589, in force since 1 July 2024 for most controllers. Opt-out detail.
• Delaware — Personal Data Privacy Act, 6 Del. C. ch. 12D, in force since 1 January 2025. Profiling detail.
• Washington — My Health My Data Act, ch. 19.373 RCW, in force since 31 March 2024, with a private right of action. Inference detail.
• California — the CCPA as amended, plus the Privacy Protection Agency’s automated decision-making technology regulations. ADMT detail.

Not law: vetoed, pending, dead

• Virginia HB 2094 — vetoed. The High-Risk Artificial Intelligence Developer and Deployer Act passed both chambers in 2025 and was vetoed by the Governor on 24 March 2025. It imposes nothing on anyone.
• Connecticut SB 2 — repeatedly not enacted. A Colorado-style high-risk AI bill passed the Connecticut Senate in 2024 and again in 2025 without clearing the House. Check the Connecticut General Assembly’s bill page for the current session before treating it as anything else. See the status page.
• New Jersey — AI hiring bills pending. Notice and bias-audit bills have been introduced across several sessions without enactment; what binds employers is the Law Against Discrimination. Detail here.
• New York — the RAISE Act. Frontier-model safety legislation passed both chambers in June 2025 and its status changed late in 2025 through the chapter-amendment process, which can alter both substance and effective date. Confirm the current position on the New York State Senate’s bill page at nysenate.gov before relying on any characterisation of it, including ours.

There is one further category that is easy to miss: a federal preemption attempt. Proposals to bar states from enforcing their own AI laws for a period of years have been floated in Congress and have not been enacted at the time of writing. If one ever is, most of the first table becomes unenforceable at a stroke, which is a reason to treat any multi-year state compliance plan as revisable.

How to verify any row yourself

The method matters more than the table, because the table ages and the method does not.

• Start at the legislature, not at a tracker. Every state publishes bill status and enrolled text. The bill page tells you whether it was signed, vetoed, or died in committee, and on what date.
• Then find the codified citation. A signed bill becomes a section of the state code, and the code is what is enforced. A bill number is a historical artefact; C.R.S. § 6-1-1701 is the thing a court reads.
• Check for later amending acts. Colorado’s delay, Utah’s 2025 narrowing and California’s SB 942 amendments were all made by separate later bills. Reading the original enrolled text alone will give you the wrong answer for all three.
• Separate the effective date from the enforcement date. Several statutes take effect on one date and become enforceable, or lose a cure period, on another. Delaware and Oregon both work this way.

Applied consistently, that method also explains the disagreement between published counts. A tracker counting every enacted bill that mentions AI will report a number in the dozens; one counting only statutes imposing cross-sector duties on developers and deployers will report a handful. Neither is wrong, and neither is useful unless it says which it is counting. This page counts the second kind, plus the employment-specific instruments that predate the current wave.