Dario Amodei has a message for the world’s democracies: get your act together on AI before someone else does it for you.
The Anthropic CEO published an essay in June titled “Policy on the AI Exponential,” laying out a framework for how democratic nations should coordinate on AI safety and regulation. The core argument is that transparency measures and voluntary commitments are no longer sufficient. Amodei wants binding rules, mandatory audits, and governments with the power to pull the plug on dangerous AI deployments.
What Amodei is actually proposing
Amodei calls for a “coalition of democracies” to coordinate AI policy, restrict access to advanced AI capabilities, and establish shared safety standards.
His proposals include mandatory third-party audits for AI models that exceed a specified computational power threshold. In English: if your AI model is powerful enough to pose real-world risks, an independent group gets to inspect it before it goes live.
