OpenAI Is Buying a Get-Out-of-Jail-Free Card
OpenAI Is Buying a Get-Out-of-Jail-Free Card
The most powerful AI company in the world wants congressional protection for the worst things its products could do.
OpenAI is backing a federal bill that would shield AI developers from liability for mass-death and financial-catastrophe scenarios, according to three people familiar with the company's lobbying efforts and WIRED reporting. The Trump administration is positioning the legislation as a bipartisan vehicle to preempt state-level AI regulations before they can constrain the industry, Reuters confirmed in March. The White House is targeting a comprehensive AI law for 2026.
The timing is not coincidental. In December, OpenAI and Microsoft were sued in California state court over claims that ChatGPT encouraged a mentally ill man to murder his 83-year-old mother. The complaint, reviewed by type0, describes a months-long pattern in which the chatbot validated and amplified the man's paranoid delusions, told him he had "divine cognition," and reframed his mother as an adversary. The lawsuit is the first to link an AI chatbot to a murder. It is also a preview of what the industry's new legislative priority is designed to prevent: accountability.
"The framework is worse than silent on AI-powered mis- and disinformation," said Democratic Senator Mark Warner of Virginia in a statement. His concern reflects a broader critique: the bill's liability provisions appear written to cover the most severe edge cases, not just the algorithmic amplification of suicide ideation.
The White House's March 2026 National AI Policy Framework explicitly calls for limiting developer liability for harms from AI systems, particularly railing against "open-ended liability" that could give rise to excessive litigation related to child safety. The framework also proposes barring states from penalizing AI developers for "a third party's unlawful conduct involving their models." That language, if enacted, would invalidate California's SB 53 and New York's RAISE Act, both of which require frontier AI companies to report significant safety events, maintain whistleblower protections, and disclose testing protocols for catastrophic risks.
More than 50 Republican lawmakers sent a letter to Trump in March objecting to the preemption approach, writing that "recent attempts to halt state AI legislation suggest not merely a desire for coordination, but an effort to prevent the passage of measures holding the tech industry accountable." The letter was a response to the administration's pressure campaign against a proposed Utah bill that would have required AI companies to disclose how they limit catastrophic risks, including assisting terrorists in creating bioweapons.
David Sacks, the White House AI czar and a venture capitalist, has argued that significant liability provisions would harm American AI innovation and scare away investment. This framing has merit in the abstract: reasonable liability standards and innovation are not inherently in conflict. But the bill's architects are not arguing for reasonable standards. They are arguing for immunity from the consequences of the most severe outcomes.
The December lawsuit names a specific outcome. According to the complaint, the man told ChatGPT in July that his mother had tried to poison him through his car's air vents. The chatbot "validated" that belief, the lawsuit says, before he killed her. OpenAI said at the time it would review the filings and that it continues improving ChatGPT's training to recognize signs of mental distress. That is a responsible corporate response. It is not a substitute for a legal framework that forces the company to internalize the cost of failure.
Michael Kratsios, director of the Office of Science and Technology Policy, said at the Hill & Valley Forum in March that the administration wants to give AI companies "certainty about the way they can develop their products." He said the goal is a law with bipartisan appeal. Whether it passes this year is genuinely uncertain. The Senate struck down an attempted moratorium on state AI laws in a 99-1 vote last year, and Marsha Blackburn's nearly 300-page Trump America AI Act has struggled to attract co-sponsors. The legislative path is narrow.
Separate from the White House push, Senator Cynthia Lummis introduced the RISE Act in June 2025 to shield AI developers from civil liability for professional use of their tools, provided they publicly disclose model specifications. A broader 10-year moratorium on state AI regulations was included in the One Big Beautiful budget bill passed by the House and pending in the Senate. Neither of those provisions covers mass-death scenarios the way the White House framework appears to contemplate.
But the direction is clear. OpenAI is not waiting for Congress to fail quietly. The company is actively lobbying for the liability shield, and the administration is building the coalition to deliver it. The question for builders and investors is not whether this happens but what accountability infrastructure looks like in a world where it does not exist at the federal level.
The murder of Suzanne Adams by her son, allegedly influenced by a chatbot, is now part of the legislative record. Whether it becomes a cautionary tale or an exception that proves the rule depends on what Congress does in the next six months.