Anthropic CEO urges government guardrails for unsafe AI systems
San Francisco-based Anthropic wants governments to stop unsafe AI models before launch, putting the city’s flagship industry under sharper scrutiny.
At Anthropic’s San Francisco headquarters, CEO Dario Amodei made a direct case for government power over the city’s most influential new industry: if an AI model is judged too dangerous, regulators should be able to block or slow its release. In a June 10 interview with ABC News and a companion essay, Amodei argued that frontier AI should face FAA-style safety testing before it reaches the public.
Amodei said the onus falls primarily on AI developers to make the technology reliable, but he wants governments to step in when independent third-party assessments find “unacceptable risks.” Under Anthropic’s proposal, auditors would examine four areas: cybersecurity, biological weapons, loss of control of AI systems, and automated R&D. He framed the model as a safety regime for systems powerful enough to demand more than voluntary promises from the companies building them in San Francisco and beyond.
The proposal also reaches beyond technical risk. Amodei said AI policy has to address job displacement, pointing to wage reinsurance, retention incentives, improvements to unemployment insurance, and longer-term support tools such as taxes on AI companies or “universal capital accounts.” Anthropic said it is backing that broader policy push with a $200 million investment in research on AI’s impact on society.
The message lands at a moment when Washington is already moving. Politico reported that President Donald Trump signed an AI oversight executive order on June 2, 2026, giving the intelligence community a larger role in model testing and requiring companies to submit powerful new models for review 30 days before public release. Amodei’s version goes further, and Politico described it as the most aggressive regulatory framework backed by a major AI CEO to date.
In the ABC interview with Linsey Davis, Amodei also revisited his earlier call for an AI “pause,” saying any real pause would require countries and AI labs to agree. He said most of his recommendations are written around the United States, but he considers them relevant globally, a reminder that the policy fight now pressing in Washington is also a San Francisco story about whether the city that built these companies is prepared to scrutinize them.
This article was produced by Prism’s automated news system from verified source data, official records, and press releases, then run through automated quality and moderation checks before publishing. The system is built and supervised by the people who set the standards it runs under. Read our full AI policy.
Did this article answer your question?
