Two years ago, Colorado planted a flag no other state had dared to raise: a comprehensive law telling companies how they had to build, test and govern the artificial intelligence systems that decide who gets a job, a loan, an apartment or a diagnosis. This spring, before that law ever took effect, Colorado pulled the flag back down.
On May 14, 2026, Gov. Jared Polis signed Senate Bill 26-189, formally repealing the Colorado Artificial Intelligence Act, SB 24-205, and replacing it with a far lighter disclosure regime that takes effect January 1, 2027. The reversal was swift and lopsided. The replacement cleared the Senate 34-1 and the House 57-6, a bipartisan margin that shows how completely the political ground had shifted beneath the nation's first landmark AI statute.
What the state gave up
When Colorado passed SB 24-205 in 2024, it borrowed the architecture of the European Union's AI Act: a risk-based regime built around "high-risk" systems, an affirmative duty to avoid "algorithmic discrimination," mandatory impact assessments, deployer risk-management programs and reporting to the attorney general.
SB 26-189 scraps nearly all of it. Gone is the duty of care to prevent discriminatory outcomes, the core obligation of the original law, along with the required impact assessments, the risk-management program mandate and the reporting requirements. Gone, too, is the upfront disclosure that would have forced businesses to explain an AI system's purpose and the "nature, source and extent" of the personal data feeding it.
In their place is a transparency framework centered on "automated decision-making technology," or ADMT, that "materially influences" a consequential decision. Companies must post a plain notice that ADMT is in use. Only if a consumer receives an adverse outcome, and asks within 30 days, must a business explain the technology's role, offer a chance to correct inaccurate data, and provide "meaningful human review... to the extent commercially reasonable." Enforcement rests exclusively with the attorney general, subject to a 60-day cure period; Coloradans get no private right to sue. But the obligations industry fought hardest, the ones dictating how models had to be governed, are the ones that vanished.
The pressure campaign
Colorado's business community, led by the Colorado Chamber of Commerce, opposed SB 24-205 from the start. Polis signed the original bill only reluctantly, warning it could "tamper innovation and deter competition in an open market" and urging lawmakers to revise it before it went live.
They kept postponing instead. First set to take effect in February 2026, the law was pushed to June 30 after Polis and five other top Colorado Democrats pressured the legislature during a 2025 special session. Polis then appointed an 18-member task force, convened on the Chamber's recommendation and meeting largely out of public view. When it released its framework in March 2026, the governor praised the group's "unanimous agreement on AI policy to protect consumers and support innovation."
State Sen. Robert Rodriguez, the Denver Democrat who sponsored both the original law and its repeal, cast the rewrite as a pragmatic settlement.
"If someone is denied housing or a job, loses their healthcare, or sees their insurance rates mysteriously skyrocket at the hands of automated technology, they deserve to know what criteria went into that decision and to have an opportunity to correct mistakes," Rodriguez said in a statement. "This bill strikes an appropriate balance of protecting consumers while not being onerous on developers or the businesses who use AI technology."
Consumer and labor advocates were more grudging. The People's Alliance for Responsible Technology, a coalition of progressive and labor groups, called itself "cautiously optimistic."
"We'll keep our eye on required disclosures to workers, patients and consumers to make sure that they're protected when AI makes important decisions about their future," said Dennis Dougherty, director of the Colorado AFL-CIO. "Coloradans deserve transparency and accountability when Big Tech affects our lives."
The Electronic Privacy Information Center was blunter, warning that the repeal strips "many important safety and testing requirements" while leaving broad industry exemptions intact.
The federal shadow
Colorado did not act alone. On December 11, 2025, the White House issued an executive order, "Ensuring a National Policy Framework for Artificial Intelligence," that singled out Colorado's law and directed the Justice Department to challenge conflicting state statutes in court. Its AI Litigation Task Force soon intervened in a lawsuit Elon Musk's xAI filed in April 2026, arguing the act unconstitutionally forced developers to embed the state's preferred viewpoints into their products. On April 27, Attorney General Phil Weiser agreed to stay enforcement, citing the pending rewrite.
The provisions the administration attacked most, the discrimination duty, the impact-assessment mandate, the risk-management program, are precisely the ones SB 26-189 removed. Yet the lesson is nuanced: federal pressure reshaped Colorado's law without preempting it. Congress has repeatedly declined to pass a national framework, and every state statute remains in force pending its own fate.
What to watch
Colorado's retreat scrambles the assumption that the EU model would define American AI regulation. The state now joins California in anchoring a lighter, disclosure-and-rights approach, even as Illinois brings a new AI disclosure statute online and Texas implements its Responsible Artificial Intelligence Governance Act. The patchwork is not shrinking; it is diverging.
The immediate action moves to Weiser's office, which must complete rulemaking before January 1, 2027, defining slippery terms like "materially influence" and the standard for "meaningful human review." Watch whether the xAI litigation is quietly mooted now that the contested provisions are gone, and whether the December order's push for a preemptive national framework gains traction in Congress. Colorado was the bellwether for AI accountability; its reversal may prove an even louder signal than its original ambition.
"This bill strikes an appropriate balance of protecting consumers while not being onerous on developers or the businesses who use AI technology."— Robert Rodriguez, Colorado Senate Majority Leader (D-Denver)