Anthropic CEO Pledges Embedded Evaluators to Slow AI

Anthropic CEO Dario Amodei has outlined three strategies to slow AI development, citing a recent OpenAI-HuggingFace hack and rapid AI capability gains as key motivators. He announced Anthropic will unilaterally invite third-party embedded evaluators to verify safety commitments, with OpenAI's Sam Altman and Elon Musk both expressing support. Amodei also called for coordination among leading democratic-nation AI companies on safety standards and proposed global cooperation, including with China, to limit dangerous AI applications. Critics remain skeptical, arguing such warnings distract from harms AI is already causing today.