Illinois SB3444 is four days from a committee vote that will determine whether the state moves forward with one of the most consequential AI liability frameworks in the country. The bill’s central mechanism is a liability shield for AI developers, protection from civil suits tied to catastrophic outcomes, provided the developer has published safety protocols meeting the bill’s requirements. Whether that shield is broad enough, or the right tool at all, is where OpenAI and Anthropic reportedly part ways.
The bill defines “critical harm” with unusual specificity. Per the bill text, the threshold sits at 100 or more deaths or injuries, or $1 billion or more in property damage. Below that threshold, the standard liability framework applies. Above it, the shield engages, but only for developers who have published safety protocols in advance. That condition matters: it’s not a blanket immunity. It’s a compliance-contingent protection.
OpenAI is lobbying in support of the bill, according to reporting from Wired and Gizmodo. The lobbying effort is reportedly connected, at least in part, to concerns about wrongful death litigation exposure, though that framing comes from journalism covering the bill, not from OpenAI’s stated position. For a company whose models are increasingly embedded in high-stakes applications, a compliance- contingent liability shield has obvious strategic value.
The picture on the other side is murkier. According to PPC Land, Anthropic is reportedly opposing SB3444 and backing a competing measure, SB 3261, which would require audited public safety plans rather than self-published protocols. That account could not be independently corroborated from higher-tier sources, and should be treated as reported, not confirmed. Still, if accurate, the split tracks cleanly onto the two companies’ public governance postures: OpenAI has generally favored frameworks that enable deployment at scale; Anthropic has consistently pushed for independent oversight mechanisms.
The April 24 deadline is reported, not confirmed from a primary legislative calendar source. But the timeline pressure is real regardless. If the bill clears committee, it moves to a floor vote with the current text, including the $1 billion / 100-casualty thresholds, which would become the operative legal standard in Illinois for frontier AI liability. If it stalls, the competing SB 3261 framework gains leverage. Either outcome shapes what liability exposure looks like for AI developers operating in the state.
For compliance and legal teams at AI companies, the immediate question isn’t which bill passes, it’s whether your organization’s existing safety documentation would qualify under SB3444’s “published safety protocols” condition. The bill doesn’t define that standard with precision in the available text. That ambiguity is a risk factor regardless of which direction Illinois goes. As the federal preemption debate continues, state-level liability frameworks like this one may be operating on borrowed time, or may end up as the template that survives federal harmonization. Watch the April 24 vote, then watch whether SB 3261 gets a hearing in the same window.