Anthropic CEO Dario Amodei has published a lengthy letter, "We Must Pace the Frontier," proposing measures to prevent AI development from outpacing safety work, drawing public support from rivals including Sam Altman, Elon Musk, Demis Hassabis and Satya Nadella. In this response, AI safety researcher Stuart Russell welcomes the intent but argues the "pacing" framing, borrowed from Formula 1's safety-car concept, is fundamentally flawed, since slowing capability progress does not itself guarantee that safety problems get solved in the time gained.

Amodei's plan has three parts: embedding third-party evaluators with full system access inside AI companies (which Anthropic will adopt unilaterally), asking frontier firms in democratic countries to agree common safety standards and limits on unchecked progress, and eventually extending a similar compact to include authoritarian states. Russell notes the letter follows a turbulent period for the industry, including the resignation of an Anthropic safety researcher and fallout from an OpenAI/Hugging Face incident, and points out that a similar "Pacing the Frontier" letter in July was signed by 1,386 AI lab employees warning of an "absence of credible plans" for controlling superintelligent systems. His central objection is that safety requirements are non-negotiable and must be tied to concrete goals, not simply to adjusting the speed of development.