Colorado Gutted Its Own AI Law. The Vote Was 57-6.
Colorado spent two years building America's first meaningful AI transparency law — one that required companies to explain how algorithms make life-altering decisions. Then they replaced it with a notification email you'll ignore. Governor Polis signs it today.
Colorado spent two years building the first meaningful AI accountability law in the United States. The original bill, SB 24-205, required companies to explain how their AI systems make consequential decisions — hiring, loans, housing, the stuff that shapes whether your life goes one direction or another. Today, Governor Jared Polis signs its replacement: a law that tells you AI was involved. Not how. Not why. Not whether it was wrong about you. Just that a machine touched the decision somewhere upstream. Two years of political effort, distilled into a notification email you’ll click past.
The original law was legitimately ambitious. It made Colorado the first state to put meaningful disclosure requirements on AI systems used in high-stakes decisions — the kind that determine whether you get the apartment, the job, or the loan. The requirement wasn’t radical: companies had to explain the logic of automated decisions affecting real people. What followed was a textbook industry pressure campaign. xAI filed a federal lawsuit claiming that AI antidiscrimination requirements violated Grok’s free speech — a legal theory aggressive enough to create a chilling effect without needing to win in court. Lobbyists from across the tech industry descended on Denver. The legislature, sensing political risk, began drafting the exit ramp. SB 26-189 — the replacement — sailed through with a 57-6 House vote and a 34-1 Senate vote, making “the tech industry rewrote the law” less of a conspiracy theory and more of a vote count.
You went from “explain how AI made the decision” to “tell users AI was involved.” That’s not regulation. That’s a notification email you’ll click past before you’ve read the subject line.
You went from “explain how AI made the decision” to “tell users AI was involved.” That’s not regulation. That’s a notification email you’ll click past before you’ve read the subject line. The legal analysis of what SB 26-189 actually requires makes this concrete: the original law put the burden on companies to document their AI systems’ decision logic, conduct bias impact assessments, and disclose those assessments to users who were affected by automated decisions. The replacement law requires notification that an AI system exists. The black box stays shut — it just has to have a label on the outside. The compliance deadline, originally June 2026, has been pushed to January 2027, giving companies eight more months to operate untouched while legislators congratulate themselves on a near-unanimous vote for consumer protection. The coverage here of voluntary NIST pre-deployment testing documented the same pattern at the federal level: the industry designs its own oversight, the government endorses the framework, and the public is told accountability has been addressed.
The proponents’ defense of the new law is that the original was too broad, would have driven a tech exodus from Colorado, and would have been litigated into irrelevance anyway. That’s plausible as a technical argument and irrelevant as a policy outcome. The question isn’t whether SB 24-205 was perfectly drafted — first-generation regulation rarely is. The question is whether the replacement does anything meaningful, and the answer is no. The story here isn’t really about the law’s details. It’s about who led the charge to gut it. Polis is a Democrat in a blue state, signing a bill that dismantled a progressive consumer protection law under industry pressure, with near-unanimous bipartisan support. That vote count isn’t a sign of good policy — it’s a sign that “tech-friendly” has become the only position that survives the lobbying process in both parties, and that Colorado’s two-year fight ended not with a compromise but with a capitulation dressed as one.
The original law was imperfect. The replacement is decorative. If the AI systems currently deciding your housing applications, your job screenings, and your loan approvals turn out to be biased, wrong, or opaque — you will now be notified. You won’t find out how, or why, or what recourse you have. You’ll just know that somewhere in the process, an algorithm was involved. Colorado built something worth building, watched the industry threaten and lobby it into dust, and then voted 57-6 to hand them the shovel.
Sources: Colorado Sun · CPR News · Axios Denver · Troutman Privacy · Colorado Sun / xAI lawsuit