There currently no shortage of open-letters that are going around in the AI space. This one is a little different from the recent open-weights letter that had NVIDIA, Microsoft and eventually OpenAI signing on while Anthropic sat out. This time, it’s employees, not companies, and the ask isn’t about openness. It’s about speed.
Over 1,100 employees from the world’s leading AI labs, including OpenAI, Anthropic, Google DeepMind and Meta, have signed a statement called “Pacing the Frontier,” asking the US government to support an international effort to develop the tools needed to deliberately slow down the pace of automated AI development. The signatories span rival companies that rarely agree publicly on anything, which is itself the point the letter is trying to make.
The statement argues that the leading AI companies believe they could be close to automating AI research itself, a moment where AI starts building better AI with less human involvement in the loop. Nobody quite knows how much this accelerates progress once it starts, but the letter’s authors say the risk is real enough that capability development could outrun the world’s ability to understand or control what it’s built. The core problem the letter identifies isn’t a lack of will inside individual labs, it’s that no single company can afford to slow down first. Each one is boxed in by competitive pressure, and each country by the other, which means nobody unilaterally pumps the brakes even if everyone privately agrees someone should.
“We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development,” the letter says.
Who’s Behind It
The list of signatories reads like a rival-lab reunion. Jakub Pachocki, OpenAI’s chief scientist, and Mark Chen, its chief research officer, signed alongside Jared Kaplan, Anthropic’s co-founder and chief science officer, and Jack Clark, the company’s head of public policy. Meta’s chief scientist Shengjia Zhao and its VP of AI research Dawn Song are on the list too, as is Anca Dragan, who leads AI safety and alignment at Google. Chris Olah, Anthropic’s interpretability lead, and John Schulman, now chief scientist at Thinking Machines, both signed as well.
What makes the list notable isn’t just the titles, it’s that these are people who compete for the same talent, the same compute, and the same customers, choosing to put their names on the same page asking for the same thing: a mechanism to pace themselves collectively, because none of them trust the incentives to let them do it alone.
What The Signatories Are Saying
The comments attached to the signatures are where the letter gets specific about what’s driving the anxiety. Leo Gao, a member of technical staff at OpenAI, compared the current moment to a nuclear chain reaction, writing that the industry is “locked in a deadly race towards an intelligence explosion” and that coordination is the only way to survive it. Matthew Rahtz at Google DeepMind, who has spent three years working on AI capability evaluations, said the recent pace of progress has been “a shock” even by his own standards, and argued that avoiding catastrophic harm from AGI will require “much, much better coordination.”
Micah Carroll, who works on misalignment preparedness at OpenAI, framed the concern in more operational terms, warning that new models arriving every few weeks raises the stakes of misuse and misalignment faster than mitigation efforts can keep pace with, and that margins for error will keep shrinking under international competitive pressure. Shengjia Zhao’s own comment on the letter echoes the same idea from Meta’s side, noting that frontier labs are close to building AI that exceeds top human performance on nearly every metric of intelligence, and that getting there responsibly matters more than getting there fast.
An Odd Fit With Some Past Comments
There’s a bit of tension worth pointing out here. Anthropic CEO Dario Amodei has previously said that AI progress essentially can’t be stopped, arguing that even if his own company halted development entirely, every other lab, and eventually China, would simply continue without them. That’s more or less the exact dynamic this letter is trying to address, not by asking any single company to stop, but by asking a government to build the scaffolding that lets everyone slow down together instead of individually losing a race nobody can opt out of.
Whether Washington is the right venue for this, or whether governments move anywhere near the speed this letter assumes they need to, is a separate question the statement doesn’t really answer. It’s also worth noting the letter comes with organizational backing from two nonprofits, Guidelight AI Standards and Encode AI, and is open for further signatures from verified employees at frontier AI companies.
For now, the letter mostly functions as a public signal, one that says a meaningful chunk of the people actually building these systems are more worried about the pace of progress than their own companies’ public messaging usually lets on.