SB 24-205, the first U.S. statute to impose broad duties on developers and deployers of high-risk AI, was supposed to take effect today. It doesn’t. Governor Jared Polis repealed it on May 14, signing SB 26-189 in its place, a thinner transparency regime that itself doesn’t take effect until January 1, 2027.

The political collapse was visible from the moment of signing. Polis put his name on the 2024 bill with explicit reservations and immediately convened an AI Policy Working Group, a tell that the law’s sponsors had won the floor vote but lost the executive branch. The duties of care and algorithmic risk assessments at the core of SB 205 were always going to be relitigated.

What finished the law was a three-front pressure campaign. x.AI filed a federal constitutional challenge in April 2026, with the Justice Department backing the suit. Trump’s December 2025 executive order on AI preemption named SB 205 by number as “excessive State regulation.” The Colorado Attorney General then joined x.AI in a joint motion to delay enforcement, an unusual posture for a state’s own enforcer.

SB 26-189 is what’s left. It covers “automated decision-making technology” that “materially influences” consequential decisions, and swaps the old duty-of-care architecture for disclosure, three-year recordkeeping, consumer rights, and fault-based liability split between developers and deployers. The AG enforces, with a 60-day cure period that sunsets January 1, 2030. Implementing rules are pending.

x.AI has signaled it’ll file a new preliminary injunction motion by June 11, aimed at the 2027 effective date. The Colorado AI Act’s real legacy may be procedural: a demonstration that comprehensive state AI regulation can be unwound before its first enforcement day by the same officials who enacted it.

Sources

Sources