Anthropic CEO Urges Binding Regulations to Block Dangerous AI Models

Anthropic CEO Dario Amodei published an essay titled "Policy on the AI Exponential" on June 10, 2026, arguing that artificial intelligence is advancing faster than government regulatory capabilities. He warned that frontier models, including Claude Mythos Preview, now present immediate risks to cybersecurity and national security, with biological and autonomy risks likely to follow.

Amodei proposes a regulatory framework modeled after the Federal Aviation Administration, requiring mandatory third-party technical testing and auditing based on a compute threshold. He advocates for granting governments the legal authority to block or reverse the deployment of models that present unacceptable risks to public safety. This proposal is more aggressive than a June 2 executive order from President Donald Trump, which focused on voluntary sharing and intelligence community testing.

To address potential enduring job displacement, Amodei suggests macroeconomic interventions including universal basic income, wage insurance, and higher capital gains taxes. Anthropic supported these calls by releasing a legislative proposal for frontier model testing, an economic policy framework, and announcing a $200 million investment to research AI's societal and economic impacts. Amodei also called for a global coalition of democracies to coordinate safety standards and prevent authoritarian regimes from using AI for mass surveillance.