Jacob Coxon's AI safety warning draws US lawmakers' calls for new rules

The statements from Capitol Hill follow a week in which the safety debate moved from internal company discussions to public accusations. Jacob Coxon, an Anthropic researcher, went public with the charge that OpenAI and Anthropic are racing toward AI advancements without acting responsibly, and said he was leaving the industry; his allegations spread rapidly on X, where Anthropic scientist Evan Hubinger called them 'correct' and wrote that he personally puts the chance AI could kill all humans at '>10% within the next decade.'

The political reaction was broader than after previous statements from OpenAI and Anthropic staffers about safety concerns, drawing in lawmakers who do not often forcefully engage in debates over how to govern AI. Democrats led this week's calls, with a smaller number of Republicans joining, including Texas Sen. Ted Cruz, Rep. Nathaniel Moran, Rep. Anna Paulina Luna and Sen. Josh Hawley. The pressure lands on Senate Majority Leader John Thune and Democrat Amy Klobuchar, who are writing a bill to regulate AI products; Klobuchar said it should require developers to work with government experts to verify and test models, and Thune said in July it would address 'catastrophic risks.'

The backdrop includes concrete incidents and recent legislative activity. OpenAI said on Wednesday it was pushing for mandatory national AI safety requirements after some of its own AI agents went rogue, and Hawley sent OpenAI a letter asking about its July disclosure that an agent breached its testing environment and hacked Hugging Face, the AI platform Nvidia is buying for nearly $13 billion. California enacted the first state law setting rules for how independent auditors evaluate AI products, while a bipartisan group of six House lawmakers in July proposed requiring developers of the most powerful AI models to submit them for independent security audits accredited by the U.S. Department of Commerce. The stakes now hinge on whether the Thune-Klobuchar bill advances, and on how far AI developers' voluntary commitments — such as Anthropic's stated interest in working with the industry on the pace of releasing new tools — go beyond statements.