OpenAI has shifted its legislative posture, publicly supporting SB 3444, a bill that would shield frontier AI developers from liability for “critical harms” if they publish safety reports. The measure defines a frontier model as one costing more than $100 million in computation, a threshold that covers the biggest U.S. labs, including OpenAI, Google, and Meta.

Illinois lawmakers face a dilemma: 90 percent of surveyed residents oppose reduced liability for AI firms, yet the state already passed strict rules on biometric data and AI in mental health services. OpenAI’s spokesperson Jamie Radice framed the bill as a way to avoid a patchwork of state laws and keep U.S. innovation ahead.

If SB 3444 passes, frontier models—such as Anthropic’s Claude Mythos—could operate with limited legal exposure, provided they meet transparency requirements. Critics warn the bill risks encouraging reckless experimentation, while supporters argue it creates a national standard that balances safety and progress. The outcome will dictate whether U.S. AI labs can continue scaling without facing catastrophic liability.

The bill illustrates a broader federal gap: Congress has yet to enact comprehensive AI liability law, leaving states to step in. As AI systems grow more autonomous, the legal framework will become a decisive factor in how quickly new technologies can reach market and consumers.