OpenAI backs Illinois bill that could shield AI labs from liability for “critical harms”

What the bill would do

It has been reported that OpenAI testified in favor of Illinois Senate Bill 3444, a proposal that would limit when AI developers can be held liable—even for so-called “critical harms” such as the death or serious injury of 100 or more people, or at least $1 billion in property damage. Under SB 3444, frontier models (defined as any model trained with more than $100 million in compute) would be exempt from liability for those extreme outcomes so long as the developer did not act intentionally or recklessly and had published safety, security and transparency reports on its website.

OpenAI’s pitch: safety plus national clarity

OpenAI framed its support as pragmatic: publish safety reports, avoid a patchwork of state laws, and aim for a national framework. A spokesperson, Jamie Radice, said the approach “focuses on what matters most: reducing the risk of serious harm…while still allowing this technology to get into the hands of the people and businesses—small and big—of Illinois.” It has been reported that Caitlin Niedermeyer of OpenAI’s Global Affairs team pushed the same line in testimony, arguing state laws should “reinforce a path toward harmonization with federal systems” and avoid friction that could harm U.S. competitiveness.

Why critics are alarmed

Not everyone is buying it. Several AI policy experts told WIRED that SB 3444 is more extreme than prior industry-backed proposals; it would, critics say, let labs off the hook for catastrophic outcomes if a disclosure checklist is ticked. Should a published report become a get-out-of-jail-free card? The tension here is raw: innovation and global leadership on one side, public safety and accountability on the other. Allegedly, the bill’s scope could cover the largest players—OpenAI, Google, Meta, Anthropic, xAI—thanks to that $100 million compute threshold.

Politics and the road ahead

Legally untested territory, politically fraught. It has been reported that some analysts think the bill has a slim chance of passing in Illinois, but its existence is significant: states are the laboratories of policy, and this one could set an industry-friendly template that pressures federal legislators. Cue the familiar Washington question: regulate now, or wait until something goes disastrously wrong? Either way, the debate over who pays for AI’s worst harms just got louder.

Sources: wired.com