Colorado passed its AI consumer protection law in May 2024. At the time, it was the first law of its kind in the United States. It required companies using AI tools to make decisions on loans, job applications, insurance, housing, and school admissions to take steps to prevent those tools from discriminating against people. It was set to take effect in February 2026, then delayed to June 30, 2026, while legislators and industry negotiated revisions.
It never took effect at all.
On April 9, 2026, Elon Musk's company xAI filed a lawsuit in federal court to block the law. The company argued, among other things, that requiring its AI chatbot Grok to avoid discriminatory outputs amounted to the government forcing the company to express views it disagreed with. Two weeks later, the US Department of Justice joined the lawsuit on xAI's side. A federal magistrate judge then froze enforcement of the law.
That DOJ intervention was not improvised. President Trump signed an executive order in December 2025 specifically naming Colorado's law as a problem. That order also created an AI Litigation Task Force whose stated purpose is to challenge state AI laws that the administration considers too restrictive. Colorado was the first target.
The DOJ's main legal argument was that Colorado's law violated the equal protection clause of the US Constitution by requiring companies to correct for discriminatory outcomes, which the DOJ framed as forcing companies to discriminate in reverse. The law's sponsors called that reading a deliberate misreading. The law required companies not to discriminate, full stop. Whether a court would have agreed is now beside the point.
Under the combined legal and political pressure, Colorado's legislature passed a replacement law, Senate Bill 26-189, on May 9, 2026. The governor signed it on May 14. The new law takes effect January 1, 2027, and is narrower in almost every direction. The requirements to test AI tools for bias, maintain risk management programs, conduct annual impact assessments, and self-report discrimination to the state attorney general are all gone. What remains is a transparency framework: companies must notify people when automated tools influenced a major decision about them, and people can request a human review.
The xAI lawsuit technically remains active. The company may still file for a preliminary injunction once Colorado completes its rulemaking under the new law.
For anyone outside Colorado who uses AI tools in their business, or whose business decisions are shaped by AI tools someone else runs, this matters for a practical reason. AI systems used in hiring, lending, insurance pricing, and healthcare already have a documented record of producing unequal outcomes across race, age, and disability status. A federal collective action lawsuit against Workday's AI hiring tools was certified in May 2025, with plaintiffs alleging the system disproportionately screened out applicants over 40. These are not hypothetical harms.
The gap that Colorado tried to close, and then was pushed to leave open, is this: existing anti-discrimination laws were written before AI existed. They cover what companies intend to do, but AI systems can produce discriminatory results without anyone intending it, because the data they were trained on reflects historical patterns of discrimination. Colorado's original law placed a duty on AI developers to test for and prevent that. The replacement law does not.
What that means practically is that the burden of proving AI discrimination still falls almost entirely on the individual who was harmed, after the fact, under laws that were not written with AI in mind. Companies that deploy AI tools to make decisions about people face no proactive obligation to check those tools are fair before turning them on.
The federal government has now made clear it will actively oppose any state that tries to change that. Thirty-plus other states are considering their own AI laws. The message sent in Colorado will reach all of them.