While the EU AI Act moves toward its 2 August 2026 milestone , the United States spent the first half of 2026 pushing in a different direction: toward one federal AI policy that would override state AI laws. A run of executive, agency, and litigation actions turned “federal preemption of a state patchwork” from a talking point into concrete moves. For anyone building AI in or for the US market, the rules you must follow, and who sets them, are actively being contested.

What happened

The federal effort advanced on four tracks, all building on Executive Order 14365 of 11 December 2025, which created a Department of Justice AI Litigation Task Force.

  • Executive Order 14409 (“Promoting Advanced Artificial Intelligence Innovation and Security”), signed 2 June 2026, is the marquee action. It directs a Treasury-led AI cybersecurity information clearinghouse, a classified benchmarking process for “covered frontier models,” and prioritised DOJ enforcement against criminal misuse of AI. It follows through on the July 2025 AI Action Plan.
  • A National AI Legislative Framework, unveiled 20 March 2026, formally asks Congress to codify a uniform federal AI policy that preempts state laws. It is a framework, not enacted law.
  • The Federal Trade Commission proposed a policy statement on 1 July 2026 arguing that, under Section 5 of the FTC Act, state laws requiring changes to “truthful outputs of AI models” are impliedly preempted where they conflict with a federal scheme. It is open for comment, not final.
  • The Department of Justice intervened in xAI v. Colorado on 24 April 2026, arguing Colorado’s AI Act (SB24-205) is preempted and violates the Equal Protection Clause. It is the first concrete court action from the new task force.

States, meanwhile, kept legislating. Colorado repealed and rewrote its first-in-nation AI Act via SB26-189 (signed 14 May 2026), resetting the effective date to 1 January 2027. Texas’s Responsible AI Governance Act took effect 1 January 2026, and California’s SB 53 transparency duties for frontier developers are in force across the period.

Why it matters for builders

The practical question for any US AI deployment is which rules bind you, and that answer is now genuinely uncertain. The federal push wants one national standard; the states have live, differing obligations on transparency, incident reporting, and algorithmic-discrimination review. Until Congress passes a preemption statute or a court settles it, both layers apply.

Two things follow. First, do not assume the state laws will be swept away: they are in force now, and the preemption framework and FTC statement are proposals, not law. Second, track the litigation, because xAI v. Colorado and any ruling on the FTC statement could reshape the landscape quickly. This contrasts sharply with the EU, where the AI Act sets one supranational baseline. Build your governance to the strictest obligation that applies to you, and revisit it as these actions resolve.

Sources

Further reading