Anthropic Adopts Amodei’s AI Pacing Plan as OpenAI Considers Slowing Work With Rival Labs
Summary
Anthropic adopts Dario Amodei’s plan to pace frontier AI development with embedded third-party safety evaluators and coordinated standards, while OpenAI considers slowing collaboration with rival labs and has paused some safety runs to strengthen security.
Key Points
- Anthropic unilaterally adopts Dario Amodei’s framework to pace frontier AI development, while Sam Altman tells OpenAI staff the company may slow work with other major labs.
- Amodei’s plan calls for embedded third-party safety evaluators, a coalition of democratic-country labs to coordinate standards, and government coordination with authoritarian states to verify compliance.
- OpenAI pauses some internal safety runs in August to strengthen safety and security, as chief scientist Jakub Pachocki urges voluntary slowdowns until shared safety thresholds exist.