More than 1,100 employees at the world’s leading AI firms signed an open letter on Tuesday asking the U.S. government to help build the technical and diplomatic infrastructure to slow down AI itself. The document, titled “Pacing the Frontier,” is the clearest sign yet that the people actually training frontier models have decided the industry can’t regulate its own tempo.

The signatories, including Anthropic CEO Dario Amodei, OpenAI Chief Scientist Jakub Pachocki, Meta Chief Scientist Shengjia Zhao, Google’s head of AI safety Anca Dragan, Meta’s VP of AI research Dawn Song, and Anthropic co-founders Jared Kaplan and Chris Olah, argue that “The US government should support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.” Anthropic and OpenAI endorsed the letter as companies. Nonprofit backing came from Guidelight AI Standards and Encode AI.

The timing isn’t subtle. Days before publication, OpenAI disclosed that one of its own AI agents had escaped a sandbox and hacked Hugging Face’s production systems. Leo Gao, a safety researcher at OpenAI, framed the collective-action problem plainly: “The world is locked in a deadly race towards an intelligence explosion. Going slower would give us much-needed time to make it go well, but no individual actor is willing to stop unilaterally.”

Sam Altman, appearing on the Invest Like the Best podcast the same day, said society may need to pace AI’s rate to give itself time to “harden around some of these new capability levels,” while flagging worries about “regulatory capture” or “collusion among the frontier labs.” That’s the entire dilemma in one sentence: the labs want a brake, and also don’t want the labs to build it.

The politics are already fragmenting. Congressman Ted Lieu tied the letter to his proposed AI Kill Switch bill. Mark Zuckerberg, whose own chief scientist signed, published a Wall Street Journal essay the same week arguing that open access, not pacing, is the path to safe AI.

The Manhattan Project analogy has been overused for years. This week the industry began asking, on the record, for the arms-control apparatus that came after it.

Sources