Anthropic pushes global AI pause plan as labor risks grow
What's next: The company plans talks with policymakers and rival labs in coming months on shared thresholds for slowing risky AI work.
Anthropic urged governments and AI developers to build a system for collectively slowing advanced AI work when models pose serious risks, arguing the idea is becoming more urgent as the technology approaches the ability to improve itself and build successors. In a blog post published Thursday, co-founder Jack Clark and Anthropic Institute lead Marina Favaro said AI could make human work far more efficient or displace it on a large scale. The company acknowledged a coordinated pause would be hard to enforce because labs and countries are racing to stay competitive. Anthropic also noted it loosened its own 2023 pledge earlier this year, saying it would no longer halt dangerous work if it lacked a significant lead over a competitor.