AI firms push legal armor
Two parallel moves this week show frontier AI firms trying to shape the legal ground before regulators do: OpenAI backed an Illinois bill that would narrow when AI companies can be sued for “critical harm,” and Elon Musk’s xAI sued Colorado to block parts of its new AI anti‑discrimination law. Both actions signal a strategy of seeking narrower statutory liability while contesting state rules that impose operational safeguards. ((wired.com), (bloomberg.com))
OpenAI and Elon Musk’s xAI picked two different states this week, but they were fighting the same battle: who writes the rules for powerful artificial intelligence before courts and regulators do. In Illinois, OpenAI backed a bill that narrows when model makers can be sued; in Colorado, xAI sued to block a law that tells artificial intelligence companies how to prevent discrimination. (wired.com, bloomberg.com) The Illinois bill is Senate Bill 3444, called the Artificial Intelligence Safety Act. It says a developer of a “frontier” model would not be liable for “critical harms” if the company did not intentionally or recklessly cause them and posted a safety protocol and transparency report on its website. (ilga.gov, wired.com) That bill uses a very high bar for both the models and the harms. A “frontier” model is one trained with more than $100 million in computing cost, and “critical harm” includes death or serious injury of 100 or more people, at least $1 billion in property damage, or the creation or use of chemical, biological, radiological, or nuclear weapons. (ilga.gov, wired.com) Wired reported that OpenAI testified in favor of the Illinois measure, which is a shift from mostly trying to block state bills outright. The company’s support matters because the bill would give large model developers a statutory shield in the rarest and most catastrophic cases, not just in ordinary product disputes. (wired.com, ilga.gov) Colorado took the opposite approach last year. Its Senate Bill 24-205, signed on May 17, 2024, requires developers and deployers of “high-risk” artificial intelligence systems to use reasonable care to prevent algorithmic discrimination in areas like employment, housing, lending, education, health care, insurance, and legal services. (leg.colorado.gov, content.leg.colorado.gov) The Colorado law is less about giant disaster scenarios and more about day-to-day decisions that shape people’s lives. It requires disclosures, documentation for impact assessments, public summaries of risk management, and notice to the attorney general and deployers when a developer learns of a credible discrimination risk. (leg.colorado.gov, content.leg.colorado.gov) xAI sued Colorado in federal court on April 9, 2026, before key parts of the law take effect on June 30, 2026. Bloomberg reported that xAI wants to block enforcement of the state’s safeguard requirements, arguing that the law unlawfully regulates the company’s systems, including Grok. (bloomberg.com, ppc.land) The complaint attacks more than one part of the statute. Reporting on the filing says xAI argues the law violates the First Amendment, the Due Process Clause, the Equal Protection Clause, and the Dormant Commerce Clause, which is the constitutional doctrine courts use when states are accused of burdening interstate commerce. (ppc.land, newsfromthestates.com) Put the two moves together and the pattern is pretty clean. When a state offers a narrow liability rule, a frontier lab tries to help write it; when a state imposes operating rules about testing, disclosure, and discrimination, a frontier lab goes to court to stop it. (wired.com, bloomberg.com, leg.colorado.gov) This fight is happening at the state level because the United States still does not have one comprehensive federal artificial intelligence law. That leaves Illinois writing one template for liability and Colorado writing another for consumer protection, while the biggest model companies try to lock in the friendlier version first. (wired.com, leg.colorado.gov, ilga.gov)
