← 返回首页

US AI law in 2026: which state changes should legal teams check

State AI laws that took effect on January 1, 2026

Multiple harm-focused AI laws entered into force on January 1, 2026 in California, including the Companion Chatbots Act, the California Artificial Intelligence Training Data Transparency Act, and the California Transparency in Frontier AI Act. California also began applying the California Artificial Intelligence Defense provision, under which it is not a defense in a harm action that the AI autonomously caused the harm to the plaintiff. The California Health Advice from AI provision requires AI technology providers to comply with pre-existing prohibitions on implying possession of a healthcare license without one. The California Real Estate Digitally Altered Images Disclosures rule requires a broker or person acting on their behalf who uses a digitally altered image in real-property advertising to disclose the alteration and provide access to the original image.

Texas enacted the Responsible AI Governance Act (TRAIGA), which also went into force on January 1, 2026. TRAIGA applies to private-sector entities that conduct business in Texas, produce a product or service used by Texas consumers, or develop or deploy an AI system in Texas.

Other states enacted targeted AI laws effective January 1, 2026, including Montana's property right in names, voices, and visual likenesses, Illinois' Human Rights Act and Consumer Protection Law, Oregon's HB 2748 prohibiting medical professional titles in an AI system's name or description, and Nevada's AI political advertising law.

California AI Transparency Act delayed to August 2, 2026

The California AI Transparency Act (Cal. Bus. & Prof. Code § 22757, et seq) extended its in-force date from January 1, 2026 to August 2, 2026. On September 2, 2026, an amendment (SB 1000) to the CAITA was delivered to Governor Newsom for signature; the Governor had until the end of September to sign or veto. If SB 1000 becomes law, it would take effect immediately as an urgency statute and would replace the CAITA's AI detection tool requirement with a "disclosure verification tool," eliminate the requirement that a covered provider offer users the option to include a manifest disclosure in AI-generated content, and require latent disclosures indicating whether a generative AI system created or altered the content. Legal teams tracking California should therefore re-check the operative text rather than rely on the original January effective date.

Texas TRAIGA: scope, prohibitions, and penalties

TRAIGA is not a true general AI law like the EU AI Act; it focuses on the most harmful uses of AI. The statute prohibits an AI system that is developed or deployed to encourage any person to physically harm themselves or others or to engage in criminal activity (§ 552.052). It also prohibits AI that infringes, restricts, or impairs a person's federal constitutional rights (§ 552.055(a)), that engages in unlawful discrimination against a protected class under federal or state law (§ 552.056(b)), that produces or distributes child pornography or unlawful deepfake videos or images (§ 552.057(1)), or that engages in explicit text-based conversations while impersonating a child under 18 (§ 552.057(2)).

TRAIGA does not offer a private right of action, but it carries penalties of not less than $10,000 per curable violation up to $200,000 for a continuing violation. The Texas Attorney General holds exclusive authority to enforce TRAIGA and may seek civil penalties, injunctive relief, attorney's fees, and reasonable court costs or investigative expenses. There is a rebuttable presumption that an entity subject to TRAIGA used "reasonable care" to comply, and a defendant is protected from liability when a third party uses the AI system for a prohibited purpose, meaning the intent in developing and distributing the system — not how it is misused — is determinative.

Colorado: from the 2024 AI Act to the ADMT Act

Colorado enacted the first general state AI law, the Colorado AI Act (CO Rev Stat §§ 6-1-1701—6-1-1707), in 2024, and Governor Jared Polis expressed concern that it would hamper innovation. Before the original act could be enforced, xAI filed suit in federal court (Case No. 1:26-cv-0151, D. Colo., April 9, 2026) seeking an injunction. On April 27, 2026, the magistrate judge granted a stay preventing enforcement of the Colorado AI Act.

A few days later, the Automated Decision-Making Technology in Consequential Decisions Act (Colorado ADMT Act) was introduced, enacted, and signed by Governor Polis on May 14, 2026, replacing and superseding the Colorado AI Act. The ADMT Act abandons the general duties of care in the prior law and sets narrower obligations on automated decision-making technology. It defines ADMT as technology that processes personal data and uses computation to generate output used to make, guide, or assist a decision about an individual, and a "consequential decision" as one relating to education, employment, housing, financial or lending services, insurance, healthcare services, or essential government services and public benefits.

Starting January 1, 2027, a developer of a covered ADMT must provide the deployer with technical documentation describing intended uses, training data categories, known limitations, and instructions for appropriate use and human review, and must notify deployers of material updates. Both developers and deployers must retain compliance records for at least three years. Consumers gain the right to request their personal data and correction of factually incorrect data, and the right to meaningful human review and reconsideration after an adverse outcome. The Attorney General enforces the act through the Colorado Consumer Protection Act; before initiating an action before January 1, 2030, the AG must give a 60-day notice and cure opportunity, and the act creates no new private right of action.

Connecticut CART Act and other 2026 developments

Connecticut Governor Ned Lamont signed the Artificial Intelligence Responsibility and Transparency (CART) Act on May 27, 2026. Its four main parts carry staged effective dates: whistleblower protections for covered employees of large frontier developers (Section 2) and automated employment-related decision technology (Sections 7-14) take effect October 1, 2026; companion chatbot requirements (Sections 4-6) take effect January 1, 2027; and transparency obligations on covered providers that generate synthetic digital content (Section 15) take effect October 1, 2026. Like Colorado's ADMT Act, the CART Act applies to developers and deployers, includes notice and transparency obligations, and does not require risk assessments. Its automated-decision regulation is limited to employment, whereas Colorado's covers consequential decisions across more domains.

On the provenance side, Utah enacted the Utah Digital Content Provenance Standards Act, in force January 1, 2027, and a separate AI Provenance Law, in force February 1, 2027. The federal backdrop also matters: a 2025 White House executive order directed evaluation of state AI laws and framed a national policy framework, though this context is background rather than a current operative rule.

Verification checklist for legal teams

Because bills were delayed, amended, or litigated, no secondary summary should substitute for the primary text. The practical workflow is to confirm, for each measure, the current statute text, its enactment and amendment status, and the operative effective date on the official legislative site before advising clients or building compliance controls.

For California, confirm the status of the CAITA and SB 1000 amendment and the January 1, 2026 effective dates of the companion, training-data, and frontier-transparency acts. For Texas, confirm TRAIGA's current text and the Attorney General's enforcement guidance. For Colorado, confirm the ADMT Act's January 1, 2027 regulatory effective date and the stayed status of the prior Colorado AI Act. For Connecticut, map each CART Act section to its specific effective date.

Reader questions

Which California AI laws took effect on January 1, 2026?

California's Companion Chatbots Act, AI Training Data Transparency Act, Transparency in Frontier AI Act, AI Defense, Health Advice from AI, and Real Estate Digitally Altered Images Disclosures took effect January 1, 2026.

What is the new effective date of the California AI Transparency Act?

The California AI Transparency Act (Cal. Bus. & Prof. Code § 22757, et seq) was extended from January 1, 2026 to August 2, 2026.

What are TRAIGA's key obligations for private-sector entities?

TRAIGA applies to private-sector entities doing business in Texas, serving Texas consumers, or developing or deploying AI in Texas, and prohibits AI used to cause harm, infringe constitutional rights, discriminate, produce deepfakes or child exploitation material, or impersonate a child in explicit conversations.

How does Colorado's ADMT Act differ from the original Colorado AI Act?

The ADMT Act replaced the 2024 Colorado AI Act's general duties of care with narrower obligations focused on consequential automated decisions, adds developer documentation duties from January 1, 2027, consumer correction and human-review rights, and three-year record retention, with no new private right of action.

What penalties apply for non-compliance?

TRAIGA imposes penalties of not less than $10,000 per curable violation up to $200,000 for a continuing violation, enforced exclusively by the Texas Attorney General. Colorado violations are deceptive trade practices under the Colorado Consumer Protection Act.

How should legal teams verify each measure?

Check the official legislative text, enactment and amendment status, and operative effective date for each state before drawing operational conclusions, because several 2026 measures were delayed, amended, or stayed.

Sources reviewed