1,178 frontier AI staff urge U.S. to build tools to slow the AI race

A July 2026 open statement signed by 1,178 employees of the leading AI labs — including chief scientists from OpenAI, Anthropic, Google DeepMind, Meta AI, and Thinking Machines — asks the U.S. government to back an international effort to build the technical and governance machinery needed to deliberately pace frontier AI development. The core worry is automation of AI research itself: signatories argue that once systems can meaningfully improve their own successors, capability gains could accelerate past humanity’s ability to understand or control them, and that no single company or country will slow down unilaterally while under competitive pressure.

The signatories frame this as a coordination problem rather than a call for any one lab to halt. Their explicit ask is for governments and industry to jointly develop mechanisms — monitoring, oversight, and the option to ‘buy time’ — before an intelligence explosion makes course correction impossible. Individual comments lean on nuclear-age analogies (splitting the atom, a runaway chain reaction) and note that even veteran capability researchers have been startled by the recent pace. Notably, the roster mixes named executives like Dario Amodei with staff who back the statement in a personal capacity, explicitly not on behalf of their employers.

The significance is less in any concrete policy proposal — there isn’t one yet — and more in the unusual spectacle of direct competitors publicly agreeing that frontier-wide pacing tools should exist. Some signatories go further, urging labs to start voluntarily designing coordination mechanisms now, ahead of any government mandate, to establish shared expectations before automated AI research forces the issue.