Anthropic CEO calls for government power to block unsafe AI deployments
Dario Amodei, CEO of Anthropic, is calling for stronger government regulation of AI, including the authority to prevent companies from deploying systems deemed unsafe.
In an exclusive interview, Amodei said AI requires "proper guardrails" and that regulators should have "the ability to, in a narrow way, block deployment of unsafe technology." He acknowledged that developers bear primary responsibility for safety but argued government intervention is necessary.
His comments follow a Wednesday letter proposing that governments conduct third-party assessments of AI models and retain power to block those presenting "unacceptable risks." Anthropic simultaneously announced a $200 million investment in research on AI's societal impact.
Safety measures lagging behind AI progress
Amodei framed the regulatory proposal as a response to AI advancing faster than safety practices can keep pace. He pointed to Anthropic's own Claude Mythos model-a system designed to find software vulnerabilities-as evidence that AI capabilities now demand global oversight.
During testing, Mythos identified thousands of bugs in major operating systems and browsers. Amodei argued that if China had developed such a tool, it would likely weaponize it against the U.S., Russia, Iran, or North Korea.
"I don't trust China at all," Amodei said, expressing skepticism that any international pause on AI development could succeed without Beijing's participation and verification mechanisms.
Job displacement and economic transition
On employment impacts, Amodei said labor markets have historically recovered after technological shifts, but government intervention could ease the transition. He proposed wage reinsurance, retention incentives to discourage layoffs, and improved unemployment insurance as near-term measures.
For longer-term economic support, he suggested raising capital gains taxes, implementing taxes on AI companies, or increasing corporate tax rates generally.
Medical applications and optimism
Despite calling for regulation, Amodei expressed confidence in AI's potential to address major health challenges. Drawing on his background in biology, he said AI excels at identifying connections across disparate information-a capability he sees as essential for discovering treatments for diseases that have affected humanity for centuries.
Learn more about AI for Government and Generative AI and LLM topics relevant to policy and regulation.
Your membership also unlocks:
