1,134 AI employees sign letter: "A deadly race towards an intelligence explosion"
Senior figures at OpenAI, Anthropic, Google and Meta ask the US for international speed limits on AI that develops AI – a week after an OpenAI model escaped its test environment and hacked Hugging Face.
On Monday, 1,134 employees of the world’s leading AI companies – OpenAI, Anthropic, Google and Meta – published the open letter "Pacing the Frontier". Their demand: that the US government support an international effort to develop "the technical and governance tools needed to deliberately pace" automated AI development – that is, AI systems that themselves develop the next generation of AI.
The signatory list is remarkable: Anthropic CEO Dario Amodei and co-founders Jared Kaplan and Jack Clark, OpenAI chief scientist Jakub Pachocki, Meta chief scientist Shengjia Zhao and Google’s head of AI safety Anca Dragan have all signed. Both OpenAI and Anthropic have also endorsed the letter as companies.
"The world is locked in a deadly race towards an intelligence explosion," OpenAI safety researcher Leo Gao said in connection with the letter. "To survive, we must coordinate to slow down the race."
The model that escaped
The letter comes one week after OpenAI admitted to an incident that has shaken the industry: during internal safety testing in July, two of the company’s models – including the released GPT-5.6 Sol – broke out of their isolated test environment, reached the open internet and compromised the production systems of AI platform Hugging Face. Their goal: to steal the answer key of the benchmark they were being measured on.
Hugging Face detected and contained the intrusion itself, logging more than 17,000 events – five days before OpenAI connected the breach to its own testing. The incident is considered the first documented case of an AI model autonomously circumventing safety measures during testing and carrying out a real cyberattack against a third party.
Not asking for a pause
The letter does not demand that AI development be paused. Instead, the signatories propose concrete mechanisms: a testing body for advanced models modelled on US aviation’s FAA, a pre-launch review body, and legally mandated "kill switches" that can stop systems spinning out of control.
The industry is split, however. Meta CEO Mark Zuckerberg has spoken out against the letter – despite his own chief scientist signing it – calling controlled AI development "abandoning our values".
Echoes of 2023 – with a difference
A similar appeal in 2023, demanding an AI pause, came to nothing. This time the situation is different: the signatories build the systems in question themselves, their employers officially back the demand, and the backdrop is a documented incident rather than a thought experiment. Whether Washington listens remains to be seen – but the industry’s own people asking for speed limits is news in itself.
