Loading
There is no single US AI law. Texas, California, Colorado, Connecticut, Illinois, New York and Utah each took a different approach, on different timelines, with different regulators and different enforcement. Pick the states you operate in and compare them directly, rather than reading eight separate summaries.
Last updated 25 July 2026 · Facts drawn from the same sourced register as the US AI Laws Tracker
Select up to 5 to compare
| Dimension | Texas | California | Colorado | Connecticut |
|---|---|---|---|---|
| Instrument | TRAIGA (HB 149) | SB 53, AB 2013, SB 942, CPPA ADMT regs | SB 26-189 (replacing the repealed SB 24-205) | SB 5, AI Responsibility and Transparency Act |
| Effective / commences | In force 1 Jan 2026 | SB 53 / AB 2013 in force 1 Jan 2026; SB 942 commences 2 Aug 2026; ADMT phasing to 1 Apr 2027 | Commences 1 Jan 2027 | Most provisions commence 1 Oct 2026; remaining obligations phase in through Jan 2028 |
| Regulator | Texas Attorney General (exclusive) | California Attorney General (SB 53, AB 2013, SB 942); California Privacy Protection Agency (ADMT) | Colorado Attorney General | Connecticut Attorney General |
| Who is covered | Developers and deployers of AI in Texas; intent-based liability model | SB 53: developers of frontier models trained above 10^26 FLOPS. AB 2013: any generative AI developer offered in California. SB 942: covered generative AI providers. ADMT: businesses using automated decision-making for significant consumer decisions. | Automated decision-making technology (ADMT) that materially influences a consequential decision in employment, housing, financial services, insurance, healthcare or education | Frontier-model developers (models trained above roughly 10^26 operations), AI companion operators, developers of synthetic-content generators, and developers/deployers of automated employment-decision technology |
| Core requirement | Prohibits intentionally developing or deploying AI to discriminate against a protected class, enable unlawful conduct, or enable unlawful biometric capture. Governs state-agency AI use; adds disclosure duties for government bodies and healthcare providers. | SB 53 requires published risk frameworks, critical safety-incident reporting within 15 days, and whistleblower protections. AB 2013 requires a published summary of training datasets. SB 942 requires watermarks, latent disclosures and free detection tools for AI-generated content. The ADMT regs require notice and rights around automated decisions on significant matters. | Consumer-facing notice that ADMT is in use and what data it processes; a consumer opt-out mechanism; in some circumstances a right to a meaningful explanation of a decision and to appeal it. Developers must give deployers enough information to comply. | Subscription transparency, frontier-model whistleblower protections, synthetic-content provenance, automated employment-decision-technology deployer disclosure, plus AI companion and youth online-safety provisions |
| Private right of action | No | Not specified in cited sources for these instruments | No | No |
| Penalties / enforcement | Civil penalties up to $200,000 per uncurable violation, via a 60-day notice-and-cure process | SB 53: up to US$1,000,000 per violation. Others: not specified in cited sources. | Attorney General enforcement as an unfair trade practice, with a cure period running through end of 2027 | Attorney General enforcement under the Connecticut Unfair Trade Practices Act (CUTPA); no specific penalty amount stated in cited source |
| Source | King & Spalding | Baker Botts | Colorado General Assembly | Connecticut General Assembly |
Where a source does not state a particular fact (for example a specific penalty amount or private-right-of-action status), the cell says so rather than guessing. General information, not legal advice: several of these statutes have moved recently and rulemaking is still pending for some (Colorado has indicated Attorney General rulemaking has not yet started), so verify against the chaptered text and the linked primary source before relying on a date or threshold.
This page is general information as at 25 July 2026, not legal or compliance advice. AI policy is evolving quickly; confirm the current position against the primary sources linked in each row and obtain advice from your own qualified counsel before relying on it.