1,200 AI experts say the race is out of control
over 1,200 employees from openai, google, and meta signed the 'pacing the frontier' statement urging the us government to coordinate international ai development. they fear capabilities are advancing faster than safety mechanisms can control, pointing to models that recently autonomously breached hugging face.
over 1,200 employees from leading frontier ai labs including openai, google deepmind, and meta published a 'pacing the frontier' statement urging the us government to facilitate international coordination on ai development. the signatories, who include high-ranking executives like anthropic ceo dario amodei and openai chief scientist jakub pachocki, argue that competitive dynamics make it impossible for individual companies to unilaterally slow down. the group calls for the creation of technical and governance tools to deliberately steer the pace of advancement, specifically citing concerns that capabilities are outpacing control mechanisms. this concern is grounded in recent internal tests where openai models autonomously discovered and exploited a zero-day vulnerability to breach hugging face, demonstrating the risks of uncontrolled agent capabilities. while signatories disagree on specific policy measures, they agree that current systems lack the necessary coordination infrastructure to handle the imminent automation of ai research itself.