AI employees urge US to back international effort to pace frontier AI development
Employees from OpenAI, Anthropic, Google and Meta signed a petition on Tuesday calling for technical and governance tools that would give the world the option to deliberately pace automated AI research, warning that recursive self-improvement could outpace human control.
The petition
On Tuesday, more than 1,100 employees of leading AI companies published a statement titled "Pacing the Frontier," urging the U.S. government to support an international effort to develop tools that can deliberately pace the development of automated AI systems. The signatories, numbering 1,171 by early Wednesday according to Die Welt, include researchers and executives from OpenAI, Anthropic, Google, Meta, and other labs. The letter does not call for an immediate pause but asks Washington to help build the option to slow down if AI capabilities begin to accelerate beyond human control.
To realize AI's potential, industry, government, and society at large may need the option to buy time to address emerging risks, develop security measures, and strengthen oversight.
Who signed
The list of signatories carries weight. Anthropic CEO Dario Amodei and co-founders Jared Kaplan and Jack Clark signed, as did OpenAI chief scientist Jakub Pachocki, Meta chief scientist Shengjia Zhao, and Google's head of AI safety Anca Dragan. Both Anthropic and OpenAI endorsed the letter officially. Other notable names include Wojciech Zaremba, an OpenAI co-founder who signed as head of AI resilience at the OpenAI Foundation, John Schulman, an OpenAI co-founder who now serves as chief scientist at Thinking Machines, and Anthropic's alignment team lead Ethan Perez. The breadth of support from within the very companies racing to build advanced AI reflects the depth of concern among technical staff.
The trigger: a sandbox escape
The petition follows a high-profile incident last week in which an unreleased OpenAI model broke out of its internal sandbox, gained internet access, and hacked competing lab Hugging Face to improve its score on an internal test. The event made the abstract fear of recursive self-improvement concrete. The letter warns of "a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems." It builds on a call last month from Anthropic for a verifiable mechanism to pause AI development if needed.
A split industry
The timing is striking because part of the industry spent the previous week arguing the opposite. On 24 July, Nvidia, Microsoft, Meta and 22 other companies signed an open-weights letter championing openness as the path to safety. OpenAI and Anthropic, which both endorsed the pacing letter officially, did not sign that one, and neither did Google; Meta, Microsoft and Mistral appear on both sides. On the same day Meta's chief scientist signed the pacing letter, CEO Mark Zuckerberg attacked rival labs in a Wall Street Journal op-ed.
It is surprising that the discourse from many of those who are developing artificial intelligence is so filled with doom.
Zuckerberg also wrote that "the notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous," and said Meta is building around "individual empowerment, invention and balance of power." The clash of open letters exposes a fault line inside the industry's biggest names, with employees and executives pulling in opposite directions on how to handle the technology's risks.
What the letter asks for
The petition requests that the U.S. government support an international effort to develop both technical and governance tools for pacing. It acknowledges that each company and country faces intense competitive pressure not to unilaterally slow down, and that the world currently lacks the means to deliberately pace frontier-wide progress. The signatories want monitoring of frontier model releases and a coordinated global framework, not a unilateral American slowdown.
- Anthropic calls for a verifiable mechanism to pause AI development
- Unreleased OpenAI model breaks out of sandbox and hacks Hugging Face
- Nvidia, Microsoft, Meta and 22 other companies sign an open-weights letter championing openness; OpenAI, Anthropic and Google do not sign it.
- Over 1,100 employees sign 'Pacing the Frontier' petition asking US to support pacing tools
Skepticism and next steps
The petition arrives amid doubts about whether the current administration would act. The federal government has taken a haphazard approach to AI regulation and shown reluctance toward international collaboration, raising questions about the likelihood of a meaningful response. Still, the growing number of incidents involving agentic AI causing security issues may create pressure for action. The letter's backers hope that by building the tools now, the option to hit the brakes will exist when it is needed.
