Trump rejects Anthropic CEO's call to slow AI

Dario Amodei runs Anthropic, one of the AI industry's most safety-focused companies. On Saturday, he published a long essay asking the whole industry to slow down anyway. He's worried that a swarm of AI programs could seize parts of the internet within six to 12 months. Using hijacked computers, it could cause hundreds of billions of dollars in damage. Sam Altman at OpenAI and Elon Musk at xAI, two of his biggest rivals, agreed within hours.

Amodei's plan has three parts. Independent evaluators would get inside access to AI labs, similar to employees. Labs would publish safety findings without a company editing them first. Actually coordinating between competing labs would need government approval, since normally that risks antitrust law. President Trump rejected the whole idea on Sunday. He said the U.S. is beating China on AI and wants to keep it that way. "Whoever wins with AI wins," he said. China's state-run Global Times fired back, calling the slowdown call a Cold War tactic aimed at holding China back.

Keep reading — what to watch

Will any lab actually let outside evaluators watch its work? Altman and Musk agreed in social media posts, not in writing or with a signed deal. King Charles is meeting AI company leaders in Scotland this week, pushing a similar idea. None of that requires any company to change what it's building right now. The real test is whether a rival lab announces its own evaluator program in the next few months. Or it just becomes something everyone quoted and nobody acted on.