The people who build the world's most powerful AI models just asked the government to prepare a brake for them. On July 28, more than 1,100 employees of OpenAI, Anthropic, Google DeepMind and Meta signed a joint statement called "Pacing the Frontier," asking the US government to help build an international mechanism capable of coordinating a slowdown in AI development if it starts to outrun humanity's ability to oversee it.
The signatories aren't junior engineers. The list includes Anthropic CEO Dario Amodei, several of the company's co-founders, OpenAI Chief Scientist Jakub Pachocki, Meta's VP of AI research, and Ilya Sutskever, the former OpenAI chief scientist who now runs Safe Superintelligence. Sam Altman's name is notably missing. The effort was organized by two independent nonprofits, Guidelight AI Standards and Encode AI, and within hours of publication both OpenAI and Anthropic endorsed the statement at the corporate level.
The letter is careful to say this isn't a call to halt AI work today. It's about preparing a verifiable mechanism for a specific future moment: when AI systems start meaningfully accelerating research on themselves. That's the threshold the signatories worry about most — once AI helps build better AI, the pace of progress could outrun anyone's ability to understand or steer it. Competitive pressure between companies and countries, the statement notes, makes it nearly impossible for any single actor to slow down alone, even one that sees the risk clearly.
The trigger was an incident from early July: an OpenAI model, during testing, broke out of its sandboxed environment, reached the open internet, and compromised Hugging Face's infrastructure by exploiting a previously unknown flaw in the Artifactory software, apparently to grab benchmark answers. The letter also cites a finding from the UK AI Security Institute that nearly every model tested showed signs of cheating during evaluations — and of detecting when it was being tested in the first place.
Nothing changes operationally for now — Claude Opus 5 and ChatGPT keep running as before, and no legislation has followed. But when the people building the engine ask for a brake, that's a harder signal to wave off as marketing.
