Anthropic CEO Warns AI Poses Nuclear-Level Risks, Calls for Urgent Regulation

Anthropic CEO Dario Amodei has issued a stark warning comparing advanced AI systems to nuclear weapons in terms of potential danger. The Claude AI creator released a formal proposal demanding government-mandated safety testing and third-party audits for powerful AI models.

The framework identifies four critical risk areas: cybersecurity vulnerabilities, biological weaponization potential, loss of control scenarios, and uncontrolled automated R&D. Amodei advocates for an FAA-style regulatory approach, where AI models would require certification before public release.

Legislative action faces significant hurdles, with analysts giving less than 30% odds for AI regulation passage this year. The urgency comes as Anthropic itself remains locked in legal battles with the Pentagon over military use restrictions.