Anthropic CEO Dario Amodei has criticized a Republican-backed proposal to block U.S. states from regulating artificial intelligence for a decade, calling it “too blunt” in a New York Times opinion piece. Instead, he urged the White House and Congress to collaborate on a federal transparency framework that would require AI companies to reveal how they assess and manage risks.

Amodei argued that the proposed 10-year moratorium comes at a time when AI technology is advancing rapidly, and without a federal strategy in place, it would leave the U.S. vulnerable. “A 10-year moratorium is far too blunt an instrument. AI is advancing too head-spinningly fast,” he wrote. He warned that such a ban could eliminate both local and national oversight, creating a regulatory vacuum.

The proposal, included in former President Donald Trump’s tax cut legislation, aims to override state-level AI rules that have recently been passed in several states. However, it has faced resistance from a bipartisan group of attorneys general who advocate for regulating high-risk AI applications. Amodei proposed that, instead of blocking state action, a national standard should require developers of advanced AI systems to conduct thorough testing, implement risk mitigation measures, and publicly disclose their safety protocols. This would ensure public awareness of how AI models are evaluated before release.

He noted that Anthropic, which is backed by Amazon, already follows these practices, and that competitors like OpenAI and Google DeepMind have also embraced similar transparency. However, he cautioned that as models grow more powerful, incentives for openness may weaken — making government-backed standards essential to maintain public trust and safety in AI development.