In an unprecedented show of unity across the AI industry, 1,367 researchers and engineers including CEOs, co-founders, and chief scientists from OpenAI, Anthropic, Google DeepMind, and other frontier AI laboratories have published an open letter warning that AI development is accelerating beyond human control.
The letter states: There is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems. It calls on the US government to support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.
Three dominant themes emerge from the personal statements of nearly a hundred signers, analyzed by Stuart Russell, a distinguished professor at UC Berkeley:
The Arms Race Dilemma. Multiple signers describe being trapped in a competitive dynamic they cannot unilaterally escape. The world is locked in a deadly race towards an intelligence explosion, one wrote. To survive, we must coordinate to slow down the race. At Davos earlier this year, both Dario Amodei of Anthropic and Demis Hassabis of Google DeepMind suggested they would halt development if competitors agreed.
No Plan for Superintelligence. A second major theme is the complete absence of credible plans for controlling superintelligent AI systems. One wrote: The AI industry has the explicit goal of building things smarter than humans. None of us yet know how to make sure these things stay under human control. This is, objectively, an insane and suicidal thing to do. An Anthropic safety researcher described their best bet as asking the superintelligent machine how to prevent it from taking over, giving this plan at best a 50% chance of working.
Real Fear Among Builders. I currently feel quite afraid of all paths I see that do not include a near-future negotiated slowdown, one signer said. Another warned: There is a very dark future ahead of us if we do not get this right. The shortest signing statement read simply: Think of all we love.
The letter intersects with concerns about recursive self-improvement, the subject of the Recursive conference in San Francisco in May and Anthropic's warning paper When AI Builds Itself in June.
