Skip to main content

Anthropic CEO Urges Slowing Down Frontier AI; Altman and Musk Agree

Anthropic CEO Calls for Slowing Frontier AI Development

Dario Amodei, CEO of Anthropic, is asking the AI industry to hit pause — or at least tap the brakes. In a recent statement, he laid out a three-step plan to slow down the race toward ever-more-powerful AI models, all built around one word: verifiability.

His proposal starts with companies bringing in on-site third-party evaluators. From there, he wants to see coordination among the U.S. and its allies. And ultimately? A global agreement. Amodei believes AI's recursive self-improvement is already underway. Within 6 to 12 months, he warns, groups of AI agents could use botnets to seize control of the internet. The fix, he argues, is to slow the pace of capability gains and pour that extra time into safety work — operational reliability, alignment, explainability, and rigorous testing.

A Dose of Realism

Amodei isn't naive about the politics. He openly acknowledges that any meaningful "speed limit" hinges on the U.S. first maintaining and expanding its technological edge over China. Only then, he suggests, could slowing down even be on the table. But here's the rub: a truly global cap can't work without China's participation. He floated a four-tier international agreement framework, yet conceded that comprehensive limits are unlikely to materialize anytime soon.

Altman and Musk Weigh In

The call didn't land in silence. OpenAI CEO Sam Altman retweeted Amodei's statement in support, and Elon Musk also voiced agreement. Their backing adds weight to a message that might otherwise be dismissed as one company's cautious stance.

A Pattern of Warnings

This isn't an isolated appeal. Earlier, Anthropic researcher Jacob Coker left the company with a stark warning: AI could potentially extinguish humanity within 10 years. From internal researcher alarms to the CEO's public push for speed limits, Anthropic — a company that built its brand around safety — is steadily moving the idea of "slowing down" from thought experiment to industry and policy agenda.

Key Points

  • Anthropic CEO Dario Amodei proposes a three-step plan to slow frontier AI, centered on third-party evaluators, U.S.-ally coordination, and global collaboration.
  • He warns that AI's recursive self-improvement has begun and that agent groups could threaten internet control within 6–12 months.
  • Amodei admits the U.S. must first maintain its tech advantage over China before any slowdown, and global limits remain unlikely in the short term.
  • Sam Altman and Elon Musk both expressed support for the call.
  • Anthropic's safety narrative gains momentum, following a former researcher's warning that AI could end humanity within a decade.