Anthropic urges global pause in high-risk AI development
The company wants governments and AI labs to set shared triggers and checks to slow research before risks become uncontrollable.
Anthropic is pushing for a slowdown or temporary pause in global development of high-risk AI, arguing that models are moving closer to the point where they could build successors to themselves. In a blog post published June 4, co-founder Jack Clark and Anthropic Institute head Marina Favaro said the world should have options to slow work on dangerous systems. They proposed that governments and AI developers agree on shared thresholds for when research should be slowed, along with verification mechanisms so no laboratory can quietly keep advancing. Anthropic acknowledged the plan would be difficult to implement amid intense industry competition, especially since the company previously loosened an internal commitment to halt research when it fell behind rivals. In the coming months, Anthropic plans to discuss the coordination effort with policymakers and other AI companies.