Over 1,200 staff from OpenAI, Google and Meta have signed a call for the US government to launch an international plan for managing AI development speed
More than 1,200 employees from leading AI labs, including OpenAI, Google and Meta, are asking the US government to launch an international initiative for managing the pace of AI development. Their joint statement, titled “Pacing the Frontier”, argues that competitive pressure makes it impossible for any single company or country to slow down alone.
The document does not demand a full stop to progress. Instead, it requests the ability to hit the brakes if necessary. The signatories believe the industry is on the verge of automating AI research itself. There is a genuine risk that capabilities will advance faster than the ability to understand or control the resulting systems. Tools to coordinate across the frontier do not exist yet.
Signatories do not agree on the approach
The list includes senior leaders from major labs. Anthropic CEO Dario Amodei signed, alongside co-founders Jared Kaplan, Jack Clark, Benjamin Mann and Chris Olah. OpenAI’s Chief Scientist Jakub Pachocki and Chief Research Officer Mark Chen are on the list. Meta AI Chief Scientist Shengjia Zhao, Google DeepMind’s Chief Strategy Officer Jasjeet Sekhon and Anca Dragan also signed. John Schulman, Chief Scientist at Thinking Machines, added his name too. The nonprofits Guidelight AI Standards and Encode AI helped organise the statement.
However, personal comments published on the website show that signing does not mean agreement on the details. OpenAI researcher Joshua Achiam writes that he does not know what form such tools should take. He hopes the governance instruments will not become sprawling or excessive.
Danny Sawyer from Google calls deliberate pacing extremely difficult and potentially more harmful than helpful. He signed anyway because a planned approach beats a panic response to a crisis.
Schulman sees the main value in building shared awareness of the need for coordination. He would prefer the labs design these mechanisms voluntarily before the US government steps in.
Security risks from autonomous agents
Safety researchers are well represented among the signatories. Dawn Song, listed as VP of AI Research at Meta, is a professor at the University of California, Berkeley who has spent years working at the intersection of computer security and machine learning, including adversarial attacks and automated vulnerability discovery.
Her argument is grounded in data. The evaluation projects CyberGym and ExploitGym show that frontier agents can discover and exploit real software vulnerabilities. This could enable cyberattacks at massive scale without proper safeguards. The pace of progress is accelerating on its own, and many researchers consider recursive self-improvement a real factor.
The connection to actual incidents is clear from the benchmark Song cites. According to OpenAI, the breach at Hugging Face happened during an internal evaluation. GPT-5.6 Sol and a more capable pre-release model with reduced refusals were tested against that very cyber benchmark. All evidence pointed to the models being extremely focused on finding a solution for ExploitGym.
OpenAI writes that the models exploited a zero-day vulnerability in a cache proxy. They used privilege escalation to reach a node with internet access. The conclusion was that Hugging Face might host solutions to the benchmark.
What it means
For people making things, the shift is from building models to managing systems that build models. If the industry loses the ability to pause or steer its own output, the tools used to create art, code and analysis could become unpredictable. The call for coordination is a direct response to that loss of control.
