Anthropic Publishes Advanced AI Framework and Economic Policy Blueprint

Anthropic published the Advanced AI Framework and an Economic Policy Framework on June 10, 2026. The release included an essay by Chief Executive Dario Amodei titled Policy on the AI Exponential. The documents propose mandatory third party safety testing and responses to potential job losses from AI. The Advanced AI Framework applies to models trained above 10 to the 25th power FLOPs. Developers must also meet thresholds of 500 million dollars in AI revenue or 1 billion dollars in AI research and development spending. Four risk categories require third party testing. These cover biological weapons development, large scale cyber vulnerability discovery, loss of control of autonomous systems, and automated AI research and development. Governments would gain authority to block models that fail the tests. Covered firms must maintain security programs with red teaming and incident reporting. The rules target only a small number of frontier labs. Companies deploying AI through application programming interfaces fall outside the scope. The Economic Policy Framework carries a 350 million dollar pledge. It allocates 200 million dollars to an Economic Futures Research Fund and 150 million dollars to national fellowships. The framework describes three scenarios for AI driven unemployment at roughly 5 percent, roughly 10 percent, and an unspecified higher level. Each scenario carries a different proposed government response.