Anthropic has released an unusually sweeping pair of AI policy blueprints that would put much tighter government controls on the most powerful systems while also reshaping how their economic gains are shared.

Early June: Anthropic publishes its frameworks

On June 10, Anthropic publicly laid out two proposals under the banner "Policy on the AI Exponential," introducing an Advanced AI Framework and an Economic Policy Framework to “prepare for AI progress.”¹ The company positioned these as proactive rules for the next generation of “frontier” AI, rather than retrofits after harms occur.

Safety framework: legal power to block dangerous AI

Reporting soon framed the move as a direct call for strong federal intervention. Axios highlighted that the Anthropic CEO says government should block dangerous AI, underscoring that the ideas go “far beyond anything currently under serious consideration in Washington right now.”²

Later that day, further details emerged: Anthropic “wants governments to have the power to block dangerous AI deployments,” giving regulators legal authority to stop models that pose a “significant risk of catastrophic harm,” with escalating civil penalties tied to a company’s global annual revenue.³ The proposal targets only the biggest players—models trained with more than 10²⁵ FLOPs and firms above high revenue or R&D thresholds—explicitly sweeping in Anthropic itself alongside OpenAI, Google DeepMind, xAI, and potentially Meta.³

Economic framework: managing disruption and sharing gains

In parallel, Anthropic’s economic framework focuses on “displacement, capital distribution, and the social safety net,” aiming to ensure AI’s benefits are “broadly shared” as the technology reshapes labor markets.³ Coverage emphasized that this second framework is designed to provide worker protections and economic safeguards as AI-driven disruption accelerates.⁴

Contrasting perspectives and implications

From Anthropic’s own materials, the message is that AI progress is exponential and requires preemptive rules to contain catastrophic risks.¹ Human commentators, meanwhile, stress how aggressive—and politically ambitious—these ideas are, noting that Anthropic is calling for regulation that would directly constrain its own business and that the proposals exceed what U.S. lawmakers are currently contemplating.²³