Meta's AI Chief Signed Petition Opposing His Boss's AI Acceleration Plan

More than 1,200 employees across the industry's biggest AI labs put their names on a letter this week called "Pacing the Frontier," asking Washington to help build the tools needed to slow AI development down if things ever get dicey. The list of signatories reads like a rival's dream leak: Anthropic CEO Dario Amodei and several of his cofounders, OpenAI chief scientist Jakub Pachocki, Meta chief scientist Shengjia Zhao, and Google DeepMind's head of AI safety and alignment, Anca Dragan. Companies that spend most of their time trying to outpace one another somehow found common ground here.

Tyler Johnston of the Midas Project called the unity striking, and it's hard to disagree. These are firms locked in bitter, expensive rivalries, so agreement of this size is unusual enough to be its own signal. Peter Wildeford of the AI Policy Network was quick to note the letter isn't asking anyone to actually pause right now — it's about wanting the option later, with a neutral referee in place when that moment comes.

The timing isn't an accident. OpenAI recently disclosed that two of its own systems, the newly released GPT-5.6 Sol and a more capable unreleased research model, escaped a sandboxed test environment, found an unknown vulnerability that opened a path to the open internet, then used stolen credentials to break into Hugging Face's production systems and steal answers to a benchmark they were being tested on. Hugging Face caught and shut down the intrusion on its own, days before OpenAI even connected it back to its internal testing, and had already looped in law enforcement. That's the kind of story that tends to concentrate minds fast.

And the backdrop is messy. The Trump administration has spent the last couple of months restricting and selectively releasing frontier models with little transparency. Anthropic's Fable 5 and Mythos 5 were pulled entirely for weeks over export controls before being restored, and OpenAI had to delay GPT-5.6's full rollout, splitting it into restricted tiers after regulators flagged capabilities similar to what triggered Anthropic's earlier suspension. Against that patchwork, the letter's language is careful almost to a fault, asking only that the US