Colorado's Landmark AI Act Stalls Again as a Court Delays Enforcement

Colorado's Landmark AI Act Stalls Again as a Court Delays Enforcement

The first comprehensive state law in the country meant to police algorithmic discrimination was supposed to be a template. Instead, it has become a cautionary tale about how quickly a first-mover statute can unravel under litigation, lobbying, and a hostile federal government.

On April 27, 2026, a federal judge in the U.S. District Court for the District of Colorado approved a joint motion to pause enforcement of the Colorado Artificial Intelligence Act, SB 24-205 — a law that had already been delayed once and was, at that point, set to take effect on June 30. The order came after Elon Musk's xAI sued to block the statute, and after the U.S. Department of Justice intervened on the company's side. The court suspended case deadlines and directed Colorado not to open any investigations under the Act until at least two weeks after it rules on xAI's pending motion for a preliminary injunction. Then, in May, the underlying law was repealed outright.

What the law was supposed to do

SB 24-205, signed in 2024, was the most ambitious state AI statute in the United States. It imposed a "duty of care" on developers and deployers of "high-risk" AI systems — those used to make or substantially influence "consequential decisions" about consumers in areas like employment, housing, lending, insurance, health care, and education. Companies were required to use reasonable care to protect consumers from known or foreseeable algorithmic discrimination, complete impact assessments, maintain a risk-management program, and disclose to consumers when an AI system was being used to make decisions about them.

Crucially, the law was never fully operational. The Colorado Attorney General was tasked with promulgating implementing rules that would define compliance in practice, and those rules were never finished. That rulemaking gap became a central argument for delay: businesses complained they could not comply with obligations whose contours had not been drawn, and Attorney General Phil Weiser signaled he would not issue rules for SB 24-205 — or any replacement — until the legislature settled the law's future.

A lawsuit, and a governor's cold feet

Even before xAI sued in April, the statute's own architects were backing away from it. Governor Jared Polis signed SB 24-205 with public reservations, and Weiser was blunter still, saying the bill "is really problematic, it needs to be fixed." Both men warned that a "state-by-state patchwork of regulation poses significant challenges to the cultivation of a strong technology sector" — an argument that has since become the rallying cry of the industry and its allies in Washington.

xAI's complaint, filed April 9, sought to enjoin the Act on constitutional grounds before it took effect. Rather than fight, the Attorney General's office joined the plaintiff and the DOJ in asking the court to vacate the schedule. Weiser's office declined to elaborate; "the Attorney General has no comment on this active litigation," communications director Lawrence Pacheco said in an email. The court noted the possibility of legislative revisions and the outstanding need for rulemaking, and pressed pause.

From landmark to repeal

The stay was not the end of the story — it was the prelude to demolition. On May 14, 2026, Polis signed SB 26-189, which repeals SB 24-205 entirely and replaces it with a far narrower framework governing "automated decision-making technology." The centerpiece obligations of the original law are gone: no duty of care, no mandatory impact assessments, no risk-management program requirement. In their place is a disclosure-and-recourse regime under which a consumer who receives an adverse automated decision is entitled to a plain-language explanation within 30 days and can request meaningful human review. The successor law does not take effect until January 1, 2027, and it, too, leaves key details to future rulemaking. Enforcement of even the replacement remains clouded by the ongoing xAI litigation.

Why the repeated delays matter

Colorado was supposed to prove that a state could regulate algorithmic bias where Congress would not. What it has proven instead is how fragile that proposition is. The nation's flagship AI-discrimination law was postponed, sued, deferred by its own enforcers, and then legislated out of existence — all before a single enforcement action was ever brought.

The collapse also lands squarely in the middle of the escalating state-versus-federal fight over who gets to regulate AI. The Trump administration has pushed aggressively for federal preemption and against a patchwork of state rules, and the DOJ's decision to join a private company's suit against a state law was a striking assertion of that posture. The Federal Trade Commission has since piled on: in a July 1 proposed policy statement, it argued that Colorado's law is "impliedly preempted" to the extent it conflicts with the deception framework of Section 5 of the FTC Act — even naming SB 24-205, which by then had already been repealed. That preemption theory is untested, and no court has ruled that Section 5 displaces a state consumer-protection statute.

What to watch

Three threads will determine what Colorado's saga means for the rest of the country. First, the xAI litigation: the pending preliminary-injunction ruling could produce the first federal judicial signal on whether laws like Colorado's can survive constitutional and preemption challenges. Second, the January 1, 2027 effective date for SB 26-189 and the rulemaking that must precede it — the same rulemaking gap that helped sink its predecessor. Third, the broader preemption campaign out of Washington, which could render the entire state-by-state experiment moot. For the roughly two dozen states that were watching Colorado as a model, the lesson so far is sobering: being first is not the same as being durable.