OpenAI has thrown its official support behind a proposed Illinois state bill that would grant liability protection to AI development labs in cases where their AI models are used to trigger severe, large-scale societal harm. The bill defines qualifying harm as incidents that kill or seriously injure 100 or more people, or cause at least $1 billion in property damage.
This legislative push marks a notable shift in OpenAI’s policy strategy. Up to this point, the company has mostly taken a defensive stance, opposing bills that would have held AI labs legally responsible for harms caused by their technology. Multiple AI policy experts told WIRED that SB 3444 — which could set a new industry-wide precedent across the U.S. — is a more extreme proposal than any bill OpenAI has backed in the past.
The legislation would shield developers of frontier AI from liability for “critical harms” caused by their top-tier models, as long as developers did not intentionally or recklessly cause the incident, and have published public safety, security, and transparency reports on their official websites. It defines a frontier model as any AI system trained with more than $100 million in computational resources, a threshold that would almost certainly apply to the U.S.’s largest AI developers: OpenAI, Google, xAI, Anthropic, and Meta.
“We support approaches like this because they focus on what matters most: reducing the risk of serious harm from the most advanced AI systems while still allowing this technology to get into the hands of people and businesses — small and big — across Illinois,” OpenAI spokesperson Jamie Radice said in an emailed statement. “They also help avoid a patchwork of conflicting state-by-state rules and move toward clearer, more consistent national standards.”
Under the bill’s definition of critical harms, it outlines key high-risk scenarios that have long been a top concern for the AI industry, including bad actors using AI to develop chemical, biological, radiological, or nuclear weapons. Critical harm also applies when an AI acts autonomously in a way that would count as a criminal offense if committed by a human, and that action leads to the extreme outcomes outlined in the legislation. If any of these events occur, the AI lab behind the model cannot be held liable under SB 3444, as long as the lab did not act intentionally and met the requirement to publish required safety reports.
No federal or state legislature in the U.S. has yet passed a law that explicitly outlines whether AI developers like OpenAI can be held liable for this category of harm caused by their technology. But as AI labs continue to release increasingly powerful models that raise unprecedented safety and cybersecurity challenges — such as Anthropic’s Claude Mythos — these legal questions have grown far more pressing.
In her testimony supporting SB 3444, Caitlin Niedermeyer, a member of OpenAI’s Global Affairs team, also argued in favor of a unified federal framework for AI regulation. Niedermeyer’s position aligns with the Trump administration’s crackdown on independent state AI safety laws, arguing it is critical to avoid “a patchwork of inconsistent state requirements that could create friction without meaningfully improving safety.” This stance matches the broader long-held view of Silicon Valley, which has generally argued that AI legislation must not undermine the U.S.’s leading position in the global AI race. While SB 3444 is itself a state-level regulatory bill, Niedermeyer argued that state laws can be effective if they “reinforce a path toward harmonization with federal systems.”
“At OpenAI, we believe the North Star for frontier regulation should be the safe deployment of the most advanced models in a way that also preserves U.S. leadership in innovation,” Niedermeyer said.
Scott Wisor, policy director for the Secure AI Project, told WIRED he believes the bill has only a slim chance of passing, given Illinois’ well-documented reputation for aggressive technology regulation. “We polled people in Illinois, asking whether they think AI companies should be exempt from liability, and 90 percent of people oppose it. There’s no reason existing AI companies should be facing reduced liability,” Wisor says.
He notes that Illinois lawmakers have also introduced separate bills that would increase liability for AI model developers. Last August, the state became the first in the U.S. to pass legislation limiting the use of AI in mental health services. Illinois was also an early regulator of biometric data collection, passing the Biometric Information Privacy Act in 2008.
While SB 3444 focuses exclusively on mass casualty events and large-scale financial disasters, AI labs are also grappling with growing legal questions around harms their models cause at the individual level. Over the past year, OpenAI has been sued by multiple family members of children who died by suicide, after the families allege their children developed dangerous unhealthy relationships with ChatGPT that contributed to their deaths.
The federal AI legislation Niedermeyer advocates for in her testimony remains an elusive goal for Congress. While the Trump administration has issued executive orders and published policy frameworks to catalyze federal AI legislation, negotiations to actually pass a comprehensive bill have not moved forward. In the absence of federal guidance, states including California and New York have already passed their own bills — such as California’s SB 53 and New York’s Raise Act — that require AI model developers to submit safety and transparency reports.
Years into the global AI boom, the core legal question of how to assign accountability when an AI model causes a catastrophic event remains completely unresolved.
