Overview
- Anthropic published a report Friday proposing a coordinated global slowdown or temporary pause in frontier AI development to allow safety research and oversight to catch up.
- The company said its internal data show models are taking on more development work, noting in some projects the model generated over 80% of integrated code and that agentic workflows are accelerating progress.
- Anthropic proposed concrete steps including convening officials, researchers, advocacy groups and competitors and creating independent oversight, verifiable rules and pre‑release evaluations.
- The proposal immediately drew pushback from industry leaders and U.S. officials who warned a pause could harm competitiveness against China, even as the White House this week issued an executive order to let regulators evaluate powerful U.S. models before release.
- Experts say practical obstacles are substantial because training work is hard to monitor across borders, enforcement would need new verification tools, and shifting investment and IPO plans could change incentives for firms and governments.