Anthropic has published two policy frameworks proposing that governments gain legal authority to block AI deployments posing catastrophic risks, with civil penalties tied to global revenue. The Advanced AI Framework targets frontier models trained above 10²⁵ FLOPs from companies with over $500M in AI revenue or $1B in R&D spend — a threshold covering Anthropic, OpenAI, Google DeepMind, xAI, and potentially Meta. Four catastrophic risk categories are identified: bioweapons, large-scale cyber vulnerabilities, loss of control over autonomous systems, and AI-automated R&D. Frontier developers would face mandatory testing, independent evaluation, and regular risk reporting. Anthropic also opposes federal preemption of state AI laws unless federal law meets the same standard. A second Economic Policy Framework addresses worker displacement through capital accounts, wage insurance, and an expanded social safety net. The move is strategically timed as Anthropic navigates a Pentagon blacklisting lawsuit, an upcoming IPO, and active Congressional negotiations on AI preemption.
