xAI sues Colorado

xAI went to federal court on April 9 and asked a judge to stop Colorado from enforcing its new artificial intelligence law before the rules kick in later this year. The company says the state is trying to regulate what an artificial intelligence model can say, not just how a business uses it. (reuters.com) The Colorado law at the center of the fight is Senate Bill 24-205, signed on May 17, 2024. It is one of the first state laws in the United States aimed at “high-risk” artificial intelligence systems used in big life decisions like jobs, housing, lending, insurance, education, health care, and legal services. (leg.colorado.gov) Colorado’s basic idea is simple: if a computer system helps decide whether you get hired, approved, insured, admitted, treated, or housed, the companies behind it should take steps to prevent discrimination. The law tells developers and deployers to use “reasonable care” against what it calls algorithmic discrimination. (coag.gov) That phrase matters because “algorithmic discrimination” in the statute means an artificial intelligence system makes a decision that unlawfully favors or harms someone based on protected traits like race, sex, disability, religion, or age. Colorado built the law around consumer protection, not around chatbots or political speech. (content.leg.colorado.gov) xAI says the law is written so broadly that it could still reach general-purpose models like Grok if those models are later used inside a covered decision. In its complaint, xAI argues that forcing a model maker to design outputs around the state’s rules is a content-based speech restriction barred by the First Amendment. (docs.reclaimthenet.org) The company is not making just one constitutional argument. The complaint also says Colorado is reaching beyond its borders by trying to control systems built and trained outside the state, and it says key terms in the law are too vague for companies to know what conduct is legal. (ppc.land) The timing is part of the pressure. Colorado’s attorney general says the law takes effect in 2026, and businesses have spent months asking for narrower language, extra exemptions, and more detailed rules before enforcement begins. (coag.gov) (hrdive.com) Colorado lawmakers wrote the bill after a year when artificial intelligence tools moved from novelty to gatekeeper. A résumé screener, a loan model, or an insurance pricing system can work like a hidden bouncer at the door, and the state wants companies to test that bouncer before it quietly locks out the wrong people. (leg.colorado.gov) Now the case turns on a question courts have barely answered: when an artificial intelligence model produces words, is that more like protected speech, more like a regulated product, or both at once. Colorado says it is policing discriminatory decision systems, while xAI says the state is stepping into the business of controlling expression. (reuters.com) (docs.reclaimthenet.org) If xAI wins, one of the toughest state-level artificial intelligence laws in the country could be blocked before it fully starts. If Colorado wins, other states will have a road map for regulating the software layer underneath hiring, lending, housing, and health decisions without waiting for Congress. (leg.colorado.gov) (reuters.com)