Anthropic CEO Dario Amodei says governments should be able to block the deployment of artificial intelligence models that do not meet agreed safety requirements. In an essay published on his website, Amodei argues that a government backstop is needed to prevent dangerous AI deployments, giving authorities the power to stop developers from releasing new models when they pose unacceptable risks. The proposal is framed as part of Anthropic’s broader emphasis on AI safety, which the company has highlighted through its approach to developing and assessing frontier AI systems. The outlets note that the position puts Anthropic at odds with some competitors, which may favor different regulatory or governance models. While the specific mechanics of how such government review or blocking authority would work are not detailed in the articles, both sources agree on the core claim: state authorities should have the ability to intervene before high-risk AI systems are deployed, provided they fail to satisfy a safety standard. The comments come as AI governance and safety oversight remain central topics in policy discussions worldwide.