1. The Colorado AI Act Died Before It Took Effect
Twelve months ago, the Colorado AI Act was the most aggressive state AI law in the country. Six weeks before its effective date, it's notice-only paperwork and a footnote in a Department of Justice (DOJ) strategy memo.
Twelve months ago, the Colorado AI Act was the most aggressive state AI law in the country. Six weeks before its effective date, it's notice-only paperwork and a footnote in a Department of Justice (DOJ) strategy memo.
The original act, signed in May 2024 with a June 30, 2026 effective date, required AI bias audits, impact assessments, and consumer disclosures backed by a $20,000-per-violation penalty schedule. It was the model other states copied. xAI sued in April 2026 on First Amendment grounds. DOJ moved to intervene on April 24, arguing the Colorado law violated First Amendment protections for algorithmic speech. Intervention is heavier than an amicus brief. It puts the federal government inside the case as a litigant, not commenting from the sidelines. On May 14, Colorado Governor Polis signed SB 26-189, replacing the original framework with notice-only requirements that take effect January 2027. The aggressive law is dead.
This was not accidental. DOJ's posture traces back to Executive Order 14365 in December 2025, which established an AI litigation task force charged with challenging state AI laws on constitutional grounds. A March 2026 Commerce Department evaluation under EO 14365 demanded federal preemption. When Congress failed to deliver (the AI moratorium provision was dropped from the FY2026 defense spending bill in negotiation), the executive branch shifted to courts. The same DOJ team will turn to California's privacy law AI amendments and New York's frontier AI bill (the RAISE Act) next. Major law firms tracking the litigation expect challenges filed within 60 days.
The federal regulatory map is moving too. The White House postponed its own AI executive order on May 21 over concerns it would slow US AI relative to China. The order, when it comes, is expected to give the National Security Agency a role in voluntary AI model testing alongside civilian agencies. Brussels delayed the EU AI Act to 2027 last week.