US AI law, in 12 questions

The United States has no comprehensive federal AI statute. Eleven tracked states write most of the rules, an executive order seeks to preempt them, and enforcement mixes attorney general suits with private class actions. These are the questions practitioners ask, answered plainly and current to October 2026. For the transatlantic view, see the EU vs US comparison.

Is there a federal AI law in the United States?

No. Congress has passed no comprehensive AI statute, and none is close. The main federal statute touching AI output is the TAKE IT DOWN Act of 2025, which makes it a crime to publish non-consensual intimate images, AI-generated ones included, and has required platforms to remove them within 48 hours of a request since May 19, 2026. The rest of the federal layer is executive action: Executive Order 14365 (December 2025) directs the Justice Department to challenge state AI laws and pressures states through funding, a June 2, 2026 order adds a voluntary 30-day pre-release vetting framework for frontier models, and agencies apply existing law, the FTC on deception, the EEOC on hiring. No comprehensive federal AI bill has passed either chamber, and none is expected to reach a floor vote before the midterm elections. Until a statute or court ruling says otherwise, the federal picture is a posture, and the operative law lives in the states.

Which US states have AI laws in 2026?

Eleven are worth tracking. Texas, Illinois, Utah, Tennessee, and Montana have AI laws in force. California, New York, and Washington have laws partially in force with more arriving, and Connecticut's CART Act, signed in May, began phasing in on October 1. Colorado's replacement law and New York's RAISE Act take effect January 1, 2027. Virginia passed a comprehensive bill and watched it die by veto. The state tracker sorts all eleven by status, effective date, and enforcement.

What is TRAIGA?

The Texas Responsible Artificial Intelligence Governance Act, effective January 1, 2026. It bans a short list of uses when they are intended to cause harm, such as inciting self-harm or crime, unlawful discrimination, and certain biometric captures, and it does not regulate high-risk categories wholesale. Only the attorney general can sue, with no private lawsuits, and civil penalties run from $10,000 to $200,000 per violation; after an attorney general win, licensing agencies can add sanctions of up to $100,000. The Texas entry has the full structure.

Is Colorado's AI Act still law?

No. SB 24-205, the first comprehensive state AI law in the country, was repealed before it ever took effect. By then xAI had sued to block it, the Justice Department had intervened, and enforcement was on hold under a stipulated order. Its replacement is narrower: an automated decision-making technology law effective January 1, 2027, focused on consequential decisions, with attorney-general enforcement, which the same court order bars for conduct through 14 days after a ruling on xAI's preliminary-injunction motion. The Colorado entry tracks passage, litigation, repeal, and replacement.

Do New York City employers still need bias audits for hiring tools?

Yes. Local Law 144 remains in force: annual independent bias audits for automated employment decision tools, published results, and advance notice to candidates. Separately, New York State enacted a frontier AI law taking effect January 1, 2027. Both live on the New York entry.

Does BIPA apply to AI systems?

Directly. The Illinois Biometric Information Privacy Act requires written consent before collecting biometric identifiers, faceprints and voiceprints included, which is exactly what many AI systems process. It carries statutory damages and a private right of action, the combination behind the largest AI-adjacent settlements in the country. A 2024 amendment limits repeated collection from the same person to a single recovery. The Illinois entry covers BIPA alongside the state's hiring and frontier laws.

Do you have to label AI-generated content in the US?

There is no general federal labeling duty. The sharpest obligation is California's AI Transparency Act, operative August 2, 2026: providers of generative AI systems publicly accessible in California, with no user threshold since September 30, 2026, must offer a free detection tool, offer users a visible disclosure option, and embed latent, machine-readable disclosures in AI image, video, and audio content. Washington's provenance law takes effect February 1, 2027, and many states regulate election deepfakes. New York has required disclosure of AI-generated performers in ads since June 2026, and California's SB 1050 does the same from January 2027. In June the Senate Judiciary Committee advanced the NO FAKES Act, which would create a federal right over voice and likeness. The transparency topic compares every regime.

What did Executive Order 14365 actually do to state AI laws?

Less than the headlines suggested, so far. It created a DOJ litigation task force and threatened federal funding. The task force's first move came in April 2026, when it joined xAI's suit against the Colorado AI Act; Colorado agreed to a court order holding off enforcement, which also covers any 2026 replacement law, and then repealed and replaced the act. The task force has not won a ruling against any state AI law, and the other state laws remain enforceable while the constitutional fight plays out. A separate September 2026 order in a private suit bars Montana from enforcing its election deepfake law against a former legislator and his political committee. Practitioners agree on the response, which is to keep complying because preemption, if it comes, may take years. The federal entry tracks the order and its deadlines.

Which state AI law has the most teeth?

Illinois, by structure. BIPA's private right of action with statutory damages produces the largest real-world exposure, and the new Safety Measures Act adds mandatory independent audits for frontier developers from 2028. California regulates the most conduct; Texas caps exposure by reserving enforcement to the attorney general. The penalty comparison puts all of them side by side.

What is the frontier-model template states are converging on?

Three states now share one design: California's SB 53, New York's RAISE Act, and Illinois' Safety Measures Act all target models trained past 10^26 operations by developers above $500 million in revenue, and all require published safety frameworks, transparency reports, and incident reporting. California and Illinois add whistleblower protections, and Illinois adds annual independent third-party audits. A de facto national standard is forming in statehouses.

What US AI rules took effect on August 2, 2026?

California's AI Transparency Act reached its operative date: large generative providers had to offer a free detection tool and a visible disclosure option and embed latent disclosures, and since September 30, 2026 the duties reach covered providers of every size. August 2 is also the date the EU's Article 50 transparency duties began to apply. The deadline calendar tracks what comes next.

Are AI hiring tools regulated in the US?

Yes, in layers. Illinois requires notice of AI use in hiring under HB 3773. New York City mandates bias audits under Local Law 144. Colorado's 2027 law covers employment as a consequential decision. California's CCPA rules on automated decisionmaking technology reach employment decisions from January 1, 2027, and its SB 947 bars discipline or termination decided by an automated system alone from July 1, 2027. And the EEOC applies longstanding discrimination law to algorithmic tools everywhere. The employment topic lines up every regime that touches hiring.