Dario Amodei called for an immediate slowdown in the pace of AI development. The Anthropic chief executive published an essay warning of harm within months. He said swarms of rogue AI agents could take over the internet in six to twelve months. He cited a recent incident involving OpenAI and Hugging Face agents. Anthropic will give outside evaluators employee-level access to check its safety procedures. Amodei asked the rest of the industry to do the same. Sam Altman of OpenAI agreed and said his company would grant similar access. Elon Musk endorsed the call. Anthropic policy chief Sarah Heck asked for a national law requiring testing of frontier models. Investor Chamath Palihapitiya said the essay also serves Anthropic's market position.
What changed
Anthropic had gone back and forth on slowing development on its own.
- 6–12 months to agent-swarm risk
- hundreds of billions in damage
Sources