Anthropic CEO Dario Amodei calls for mandatory safety testing for frontier AI models

Anthropic CEO Dario Amodei released a landmark essay titled "Policy on the AI Exponential" on June 11, 2026, arguing that the current pace of AI progress—exponential and far outstripping policy processes—demands a radically new approach to regulation. Amodei calls for mandatory third-party safety testing for frontier AI models, with government authority to block deployment of systems deemed too dangerous. He draws parallels to the FAA for aviation safety and proposes an accelerated FDA-like approval pathway for AI-discovered drugs. The essay marks a significant departure from Anthropic's previous advocacy for voluntary transparency, as Amodei now insists that self-regulation is no longer viable.

The proposal has sparked intense debate across tech and policy circles. Supporters argue that as AI scales toward what Amodei calls "a country of geniuses in a datacenter," proactive regulation is essential to prevent catastrophic outcomes. Critics, however, warn that giving governments power to block AI could be abused, with some comparing it to historical power grabs. The reaction on social media ranges from cautious support to sharp accusations of regulatory capture. Regardless, the essay moves the Overton window on AI regulation, forcing enterprises to consider new compliance regimes that could reshape how advanced AI models are developed and deployed globally.