Anthropic’s Amodei urges a careful, slower AI race for safety

Anthropic CEO Dario Amodei argues that AI development should slow to prioritize safety, warning that self-improving systems could outrun our ability to govern them. Citing the July OpenAI–Hugging Face incident as a wake-up call, he calls for ongoing outside evaluators embedded within frontier firms (Anthropic plans to do this), potential U.S. antitrust waivers to coordinate safety standards, and international cooperation to prevent a race to the top. He also notes the vast benefits of AI—like medical breakthroughs—but emphasizes that safe progress must guide advancement.