Tom Cedoz

Tracker · AI & Emerging Risk

State AI Laws: In Force and In Motion

Plenty of trackers will tell you everything moving in fifty statehouses. This page answers the narrower question that actually lands on a desk: which enacted laws, city ordinances, and agency guidance bind an employer or business deploying AI, what does each one ask, and when did someone last check. Binding instruments first; a short watchlist for what is genuinely in motion; a changelog so you can see what moved.

Verified August 15, 2026· 14 entries · 8 in motion· Prints to 3 pages

This page tracks enacted AI law with an employer’s lens — plus one city ordinance and one agency guidance with teeth: what each asks, whom it reaches, and when it was last checked against the primary source. Government-procurement rules, deepfake crimes, and likeness statutes are deliberately absent — so is every bill that is merely pending. Four tags: in force, phasing or effective on a date certain, delayed, and guidance.

California

Civil Rights Department automated-decision rules · Cal. Code Regs. tit. 2 (FEHA regulations) In force 10/1/25

Discrimination rules for automated-decision systems in hiring and personnel moves, plus a four-year recordkeeping duty. The quiet one most California employers have not read.

Reaches: California employers with five or more employees · verified 8/14/26 · regulation text

Training-data transparency (AB 2013) · Cal. Civ. Code § 3111 In force 1/1/26

Developers must post a public summary of what their models trained on. You will not file one — but your vendors should have. Ask where it is.

Reaches: Developers of generative AI made available in California · verified 8/14/26 · bill text

California AI Transparency Act (SB 942) · Cal. Bus. & Prof. Code §§ 22757–22757.6, as amended by AB 853 In force 8/2/26

Detection tools and provenance marks for AI-generated content. AB 853 moved the operative date once, from January 1, 2026 to August 2, 2026. Covered-provider duties are live; platform and hosting duties phase in January 1, 2027, capture devices January 1, 2028.

Reaches: Large generative-AI providers first; platform and device duties phase in through 2028 · verified 8/15/26 · current code

Transparency in Frontier AI Act (SB 53) · Cal. Bus. & Prof. Code § 22757.10 et seq. In force 1/1/26

Safety frameworks and incident reporting for the biggest labs. You are almost certainly not covered — and now you can say exactly why.

Reaches: Frontier-model developers above the 10^26 compute threshold; additional duties if the developer (with affiliates) had $500M+ prior-year revenue · verified 8/15/26 · bill text

Colorado

Colorado AI Act, revised (SB 24-205 + SB 26-189) · C.R.S. § 6-1-1701 et seq. Delayed — eff. 1/1/27

The May 2026 rewrite (SB 26-189, signed 5/14/26) replaces the 2024 scheme with an ADMT statute operative January 1, 2027: developer documentation, three-year records, pre-use notice, a 30-day adverse-outcome explanation, correction, and commercially reasonable human review on request. The 2024 duty of care, risk program, and impact assessment are gone. The April stay, not the May signature, is what blocked enforcement of the old act.

Reaches: Deployers of automated decision-making technology that materially influences consequential decisions, including employment · verified 8/15/26 · SB 26-189

Insurance: unfair-discrimination testing (SB 21-169) · C.R.S. § 10-3-1104.9 and Division of Insurance rules In force (phased since 2023)

If you are an insurance licensee anywhere, assume your regulator expects a written AI governance program. Colorado wrote the testing rules; the NAIC bulletin spread the expectation.

Reaches: Colorado-licensed insurers. The NAIC model bulletin has spread a similar governance expectation to other insurance departments. · verified 8/15/26 · statute

Connecticut

An Act Concerning Online Safety (SB 5) · Public Act 26-15, signed 5/27/26 Phasing from 10/1/26

From October 1, 2026, a Connecticut WARN notice must say whether the reduction relates to the employer’s use of AI or another technological change, and using an automated tool is not a defense to a discrimination claim. The employment-technology notice duties reach systems deployed on or after October 1, 2027.

Reaches: Employers doing business in Connecticut that deploy automated employment-decision technology; consumer and platform provisions reach further · verified 8/15/26 · public act

Illinois

Human Rights Act AI amendment (HB 3773) · 775 ILCS 5/2-102(L), P.A. 103-0804 In force 1/1/26

Use AI in any employment decision and you own the outcome under ordinary discrimination law, plus a notice duty to an employee. Treating applicants the same is the prudent read, not the statutory text. No bias-audit mandate.

Reaches: Illinois employers under the Human Rights Act (one or more employees) · verified 8/15/26 · public act

AI Video Interview Act · 820 ILCS 42 In force 1/1/20

Consent, an explanation, and deletion rights before AI screens a recorded interview. The oldest law on this page, and still the one most often missed in hiring stacks.

Reaches: Employers using AI to evaluate recorded video interviews for Illinois-based roles · verified 8/14/26 · public act

New Jersey

Attorney General guidance on algorithmic discrimination · N.J. Division on Civil Rights (Jan. 2025) Guidance 1/25

No new statute — that is the point. The guidance says the Law Against Discrimination already reaches AI-driven decisions. New Jersey did not wait for a bill, and neither should your compliance review.

Reaches: All New Jersey employers · verified 8/15/26 · January 2025 guidance

New York

RAISE Act, as amended · S6953-B (2025), chapter amendment 3/27/26 Eff. 1/1/27

Safety plans and incident reporting for the biggest labs — not for companies that merely use AI. Its main effect on you is what it will make your vendors disclose.

Reaches: Two tiers: frontier developers above the 10^26-FLOP threshold (transparency reports and 72-hour incident reports); large frontier developers, $500M+ prior-year revenue, take the framework, DFS disclosure, and assessment duties · verified 8/15/26 · chapter amendment

New York City

Local Law 144: automated employment decision tools · N.Y.C. Admin. Code § 20-870 et seq. — municipal, not state law In force 7/5/23

The only in-force municipal independent bias-audit mandate for employment AEDTs: independent audit, published results, and candidate notice before the tool screens anyone. Not a state law, and not the only testing duty in the country.

Reaches: Employers and employment agencies using AEDTs to screen candidates or employees for New York City roles · verified 8/15/26 · DCWP

Texas

Texas Responsible AI Governance Act (HB 149) · HB 149 (89th Leg., 2025) In force 1/1/26

A prohibited-practices list keyed to intent — discrimination, manipulation, rights violations — rather than Colorado-style duties. Lighter on employers, but intent gets proved with your documents, so write them accordingly.

Reaches: Developers and deployers doing business in Texas; attorney-general enforcement with a 60-day cure period · verified 8/14/26 · bill record

Utah

AI Policy Act (SB 149, as amended) · Utah Code § 13-77-101 et seq. In force 5/1/24

Disclosure duties live in Chapter 77 (SB 226, effective May 7, 2025): on request for ordinary consumer use; up front for high-risk regulated-occupation interactions. Chapter 72 is the Office of AI Policy, not the disclosure statute.

Reaches: Businesses using generative AI with Utah consumers; stricter rules for regulated occupations and mental-health chatbots · verified 8/15/26 · statute

In motion

The short list of what could change this page — not every pending bill, just the ones with a realistic path to binding an employer. Each line was checked on the sweep date above.

  • Colorado — The April stay, not the May 14 signature, is what blocked SB 24-205. The repeal-and-reenact takes effect January 1, 2027. Rulemaking and xAI v. Weiser decide what that date actually looks like. revision signed 5/14/26
  • Federal — A December 2025 executive order points agencies at state AI laws, and DOJ intervened in the Colorado case in April. The thing to watch is a congressional AI standard with a preemption clause. DOJ intervention 4/24/26
  • California — AB 1018 has sat on the Senate inactive file since September 13, 2025. August 31 is the last day either house may pass it; adjournment sine die is November 30. It can still be taken up; it is not moving. inactive file since 9/13/25
  • Illinois — HB 3773 is unchanged. IDHR’s proposed Subpart J notice rules were published in May 2026 and withdrawn June 2, 2026. The statute is in force without final implementing rules. Subpart J withdrawn 6/2/26
  • New York — Comprehensive bills keep arriving behind the RAISE Act; none has passed. monitoring
  • Virginia — HB 2094 died by veto in 2025 and the 2026 session produced no comprehensive act; a new administration took office in January. Watch 2027. no 2026 act
  • Massachusetts — Perennial algorithmic-accountability bills, none enacted. Formal sessions ended July 31; conference bills can still pass in informal session through year-end. formal session ended 7/31/26
  • Michigan — Bills introduced, none enacted. The employer exposure today is ELCRA and ordinary discrimination law, not an AI statute. monitoring

What changed

  • 8/15/26 — Second adversarial pass: SB 942 and RAISE pointed at current code / S8828; Utah disclosure moved to Ch. 77; CT official title and WARN wording; IL Subpart J withdrawn; AB 1018 Aug. 31 is last day to pass, not sine die; NYC scoped as municipal AEDT audit.
  • 8/15/26 — Adversarial pass: SB 942 delayed once, not twice; RAISE $500M is the large-frontier tier; Colorado 2024 act stayed then repealed-and-replaced; NJ guidance PDF; CT public act; AB 1018 on the inactive file; count recast as entries; NAIC and NYC audit hedges.
  • 8/14/26 — Tracker launched. Every row verified against the statute, session record, or agency source.
The federal backdrop

There is no comprehensive federal AI statute, and nothing on this page is preempted today. There is a campaign in that direction — a December 2025 executive order directs agencies to challenge state AI laws, and in April 2026 the DOJ intervened on the challenger’s side against Colorado. Until a court or Congress says otherwise, plan against the law that exists, not the preemption that might arrive.

Use note

Built for in-house counsel and the team deciding whether an AI deployment needs a compliance workstream. It is general information, not legal advice, and a tracker is a map, not the territory — read the statute before you rely on a row, and treat the verified dates as the promise that someone recently did.

If your state is not here, that means no employer-facing AI statute or listed guidance was in force at the last sweep — not that AI use is unregulated. Discrimination, privacy, and consumer-protection law never left. For what to do about the tools themselves, start with the governance inventory and the hiring risk review.