AI firms pick fights
The big players in AI are moving fast from product launches to courtroom and legislative fights — one company is trying to stop state rules, and another is backing limits on liability. Elon Musk’s xAI sued Colorado to block a new state AI law it says should be handled at the federal level and that it believes infringes its First Amendment rights (reuters.com). At the same time, OpenAI publicly supported an Illinois bill that would narrow when AI companies can be held liable, even in cases of “critical harm,” signalling the industry’s preference for liability limits as regulation spreads (wired.com).
Artificial intelligence companies spent 2023 and 2024 racing to launch chatbots. In April 2026, two of the biggest fights moved to courthouses and state capitols instead. (reuters.com) (wired.com) Elon Musk’s xAI sued Colorado in federal court on April 9, 2026, asking a judge to block Senate Bill 24-205 before the law takes effect. xAI says Washington, not 50 separate states, should write the main rules for advanced artificial intelligence. (reuters.com) Colorado’s law is not a general chatbot law. It targets “high-risk” systems used in consequential decisions like jobs, housing, lending, health care, insurance, and legal services, where one bad model can work like a faulty scoring system for thousands of people at once. (leg.colorado.gov) The law says developers and deployers must use “reasonable care” to reduce algorithmic discrimination. It also requires disclosures, impact assessments, notices to consumers, and a way for a person to appeal an adverse decision through human review if that is technically feasible. (leg.colorado.gov) xAI argues those rules burden speech because modern artificial intelligence systems are built from text, weights, and outputs that the company says deserve First Amendment protection. The suit also argues that state-by-state rules will fracture the market, because one model can be downloaded and used across the country in seconds. (reuters.com) Colorado is already reworking parts of the law before full enforcement begins, which shows how unsettled this field still is even after the bill was signed in 2024. The original statute says the requirements begin on February 1, 2026, while Reuters reported xAI’s suit as a bid to stop enforcement ahead of a June 30, 2026 milestone in the state’s rollout. (leg.colorado.gov) (forbes.com) (reuters.com) At the same time in Illinois, OpenAI backed Senate Bill 3444, which moves in the opposite direction. Instead of blocking a state rule outright, OpenAI supported a bill that would make it harder to sue frontier model developers when their systems are tied to extreme harms. (wired.com) (ilga.gov) The Illinois bill says a developer of a frontier artificial intelligence model would not be liable for “critical harms” if the developer did not intentionally or recklessly cause the harm and if it published a safety and security protocol plus a transparency report. In other words, the bill swaps a broad duty to pay damages for a narrower rule tied to extreme misconduct and paperwork compliance. (ilga.gov) (wired.com) Wired reported that the bill’s definition of “critical harm” includes events like mass casualties or enormous financial damage. That means the argument is no longer just about chatbot mistakes or defamation suits; it is about who pays if a powerful model is linked to something closer to an industrial disaster. (wired.com) Put those two moves together and the industry’s playbook comes into focus. When a state writes operating rules, companies argue the federal government should preempt them; when a state offers liability limits, companies show up to support the bill. (reuters.com) (wired.com) That is a familiar strategy from other industries: fight the rules that tell you how to build the product, and back the rules that narrow who can sue after something goes wrong. Artificial intelligence policy in the United States is starting to look less like a lab race and more like a long insurance and federalism fight. (reuters.com) (wired.com)
