Frontier AI: 1,178 industry workers want the option to slow down later
OpenAI and Anthropic back a letter urging governments to build the tools that could, if ever needed, pace the most advanced AI.
A rare move is shaking the AI industry. In late July 2026, a statement titled Pacing the Frontier gathered the signatures of 1,178 employees from leading AI companies, including OpenAI, Anthropic, Google and Meta. Notably, OpenAI and Anthropic endorsed it as companies within hours.
The ask, in plain terms
Signatories are not calling for an immediate stop. Their request is subtler: that governments, the United States first, help build the technical and governance instruments that would make it possible, if the day comes, to deliberately slow the race toward the most advanced AI. In other words, build the ability to brake later rather than discover too late that you cannot.
The text flags a specific risk: the automation of AI research itself, meaning systems that accelerate the building of ever more capable AI, to the point where the pace outstrips humans’ ability to understand and oversee it.
High-profile signatories
This is not an anonymous petition. The names include prominent leaders and scientists such as Dario Amodei, CEO of Anthropic, Jakub Pachocki and Mark Chen at OpenAI, Anca Dragan for AI safety at Google, and Shengjia Zhao at Meta. Support from people at the very top gives the effort unusual weight.
Why it matters
The letter caps a July in which AI safety moved from talk to practice: technical alliances, insider petitions, the first verifiable governance tools. The debate is no longer only about what to build, but how fast, and with what safeguards.
The hard question remains: how do you coordinate a possible slowdown internationally without one player using it to pull ahead? That is exactly the problem the signatories are asking governments to help solve.