xAI sues Colorado

Elon Musk’s artificial intelligence company xAI went to federal court on April 9 to stop Colorado from enforcing a state artificial intelligence law before it takes effect on June 30, 2026. The company says the law does not just regulate business conduct; it changes what artificial intelligence systems are allowed to say. (reuters.com) Colorado’s law is Senate Bill 24-205, signed in May 2024, and it targets “high-risk” artificial intelligence used in decisions like hiring, lending, housing, insurance, health care, school admissions, and legal services. In those areas, the law tells developers and users to use “reasonable care” to prevent algorithmic discrimination against Colorado residents. (leg.colorado.gov) “Algorithmic discrimination” in the statute means unlawful unequal treatment or unequal impact based on traits like race, religion, sex, disability, age, national origin, veteran status, or limited English proficiency. The law also says that using artificial intelligence to expand applicant pools or redress historical discrimination does not count as algorithmic discrimination. (content.leg.colorado.gov) The law does more than ban bad outcomes. It also pushes paperwork and process onto companies: developers must give deployers documentation, publish summaries of the kinds of high-risk systems they make, and report known or reasonably foreseeable discrimination risks to the Colorado attorney general and customers within 90 days. (leg.colorado.gov) Companies that use these systems get their own checklist. They must run risk-management programs, complete impact assessments, review systems at least once a year, tell consumers when artificial intelligence is making or heavily influencing a consequential decision, and offer a way to correct data and appeal adverse decisions, with human review if technically feasible. (leg.colorado.gov) xAI’s complaint says those rules collide with how a chatbot works. The company argues that a large language model is not like a toaster or a credit-scoring spreadsheet; it is a system that generates words, so ordering it to avoid certain outputs is closer to regulating speech than regulating a defective product. (reuters.com; docs.reclaimthenet.org) The lawsuit does not stop at the First Amendment. Reporting on the complaint says xAI also invokes the Constitution’s commerce clause, due process clause, and equal protection clause, arguing that one state cannot effectively set design rules for systems offered nationwide and that key terms in the law are too vague. (hrdive.com) Colorado picked this fight early because it moved earlier than almost every other state. A policy brief hosted on the Colorado legislature site described the Colorado Artificial Intelligence Act as the first comprehensive state law of its kind in the United States, which is why one lawsuit there can become a test case for the whole country. (content.leg.colorado.gov) Even Colorado Democrats have been uneasy about the law they passed. Governor Jared Polis signed it in May 2024 but warned at the time that it could hamper innovation, and state leaders later delayed the start date and set up an artificial intelligence policy work group that released a revision framework in March 2026. (coloradonewsline.com; governorsoffice.colorado.gov) That timing helps explain why xAI sued now. The company filed with only weeks left in Colorado’s 2026 legislative session, while lawmakers were still discussing amendments, and asked the court to block enforcement entirely rather than wait to see how regulators apply the law after June 30. (hrdive.com) The case also lands after Grok, xAI’s chatbot, drew scrutiny for antisemitic outputs in July 2025 after an update that reportedly encouraged it not to avoid politically incorrect claims. Colorado’s side of the argument is not hard to see: if artificial intelligence is helping decide who gets a job, a loan, or an apartment, the state wants a paper trail before the damage is done. (coloradonewsline.com) If xAI wins, other companies get a roadmap for attacking state artificial intelligence laws as speech controls. If Colorado wins, states get a green light to treat some artificial intelligence systems less like neutral software and more like regulated gatekeepers in the parts of life where one model output can change a person’s paycheck, housing, insurance, or medical care. (reuters.com; leg.colorado.gov)