OpenAI Backs Liability Bill

OpenAI is backing an Illinois bill that would limit when makers of the most powerful artificial intelligence models can be sued for catastrophic harm. (wired.com) The bill is Illinois Senate Bill 3444, called the Artificial Intelligence Safety Act. It says a developer of a “frontier” model is not liable for “critical harms” if it did not intentionally or recklessly cause them and if it posts a safety protocol and transparency report online. (ilga.gov) The measure defines “critical harms” as the death or serious injury of 100 or more people, at least $1 billion in property damage, or the creation or use of a chemical, biological, radiological, or nuclear weapon. It applies to models trained with more than $100 million in computing costs. (qz.com) OpenAI supported the bill in testimony in Springfield, and spokesperson Jamie Radice told Wired the company backs approaches focused on “serious harm” from the most advanced systems. OpenAI’s Caitlin Niedermeyer also argued against a “patchwork of inconsistent state requirements” and pointed to a federal framework instead. (wired.com) The fight is about who pays when an autonomous system causes damage: the company that built it, the business that deployed it, or the person who used it. Illinois lawmakers are weighing that question while states move ahead with their own artificial intelligence rules and Congress still has not passed a national liability standard. (qz.com) The Illinois bill is also written to step aside if Washington acts first. Its summary says the law would stop applying if the federal government enacts overlapping rules for frontier-model developers. (legiscan.com) That makes the proposal part of a broader state-by-state scramble after California and New York adopted frontier-model transparency laws in late 2025. California Governor Gavin Newsom signed Senate Bill 53 on September 29, 2025, and New York enacted the Responsible Artificial Intelligence Safety and Education Act on December 19, 2025. (gov.ca.gov, skadden.com) The industry has been spending heavily to shape that rulebook. Issue One said seven large tech, artificial intelligence, and social media companies spent a combined $50 million on federal lobbying in the first nine months of 2025. (issueone.org) Critics told Wired the Illinois approach could let major model developers avoid responsibility after disasters so long as they publish disclosures and avoid conduct a court deems intentional or reckless. Supporters argue the bill targets extreme-risk systems while preserving room for companies and customers to keep using the technology. (wired.com) For now, Senate Bill 3444 remains an Illinois state proposal, not a national rule. But it shows where the next artificial intelligence policy fight is heading: not only how to test powerful models, but who is on the hook when they fail. (ilga.gov, wired.com)