What would it take to put a speed limit on frontier AI?
OpenAI CEO Sam Altman endorsed “pacing” frontier development so increasingly capable models do not outrun humanity’s ability to control them, echoing Anthropic CEO Dario Amodei.[3] Amodei’s proposed sequence starts with deeply empowered independent evaluators, expands to industry cooperation and ultimately seeks an international agreement, while stopping short of calling for AI research to halt.[6]
Why it matters
Public agreement among competing AI leaders does not itself create a slowdown: the companies have not yet settled its terms, some coordination may require government support, and international verification remains difficult because developers and countries have strong incentives to pull ahead.[4][6]
Key insights
- Altman said OpenAI would welcome independent evaluators with access comparable to that of top employees and indicated that pacing has been discussed internally.[6]
- Hinton advocated safeguards including mandatory pre-release chatbot testing, while saying AI development should not stop altogether.[4]
- Amodei framed pacing as leaving more time between major capability advances so developers can evaluate models and prepare safeguards.[6]
- US President Donald Trump rejected slowing down, citing competition with China, while Amodei described enforceable international coordination as a dilemma that could take years and might not be possible.[4]