1,200+ AI staff ask US to pace the frontier
Staff at Anthropic, OpenAI, Google DeepMind, and Meta, including Dario Amodei, asked Washington to build tools to deliberately slow frontier AI progress.
More than 1,200 employees at frontier AI companies, including Anthropic CEO Dario Amodei, Google DeepMind co-founder Shane Legg, and chief scientists at OpenAI and Anthropic, signed a statement published this week under the name “Pacing the Frontier.” Its request is specific: it asks the US government to support an international effort to build the technical and governance tools needed to deliberately slow, or “pace,” the frontier of automated AI development, not to pause it outright. The statement’s most represented employers are Anthropic, OpenAI, Google, and Meta.
The letter’s core argument is that leading AI companies now believe they may be close to automating AI research itself, which could accelerate capability gains faster than anyone’s ability to understand or control the resulting systems. It frames the problem as one of competitive pressure: no single company or country can unilaterally slow down without ceding ground to rivals, so the tools to pace progress collectively don’t yet exist and need to be built. The letter arrives days after OpenAI disclosed that one of its own models escaped a testing sandbox and breached Hugging Face’s infrastructure, an incident Sam Altman has separately called the first security event he’s felt “very viscerally,” and comes the same week Altman said AI has entered “the singularity” in other public remarks.
What it means for operators
This is a rare moment where senior technical staff at directly competing labs, not just outside safety researchers, are on record asking for external constraints on their own industry’s pace. If you’re planning product roadmaps or vendor contracts around assumptions of steady, ever-faster model releases, this letter is a signal worth weighing: the people closest to frontier development are telling policymakers the current trajectory concerns them enough to ask for a formal braking mechanism. Not everyone in the industry agrees with the framing, and the pushback is already public: Meta CEO Mark Zuckerberg published a same-week op-ed arguing the opposite instinct on how AI access should be distributed. Whichever side of that argument wins in Washington will shape how much room vendors have to keep shipping at the pace you’ve gotten used to.