This month, the U.S. government effectively compelled Anthropic, a leading American artificial intelligence company, to withdraw its most advanced AI model, Mythos, from the market on very short notice. Reports indicate the company was given approximately 90 minutes to comply with the directive, an action that highlighted growing tensions between government authorities and AI developers.
Anthropic produces Mythos, a frontier AI model, and its commercial variant, Fable. The government's sudden intervention came amid broader concerns over the security risks posed by advanced AI systems. According to testimony from Gen. Joshua Rudd, head of the National Security Agency and U.S. Cyber Command, Mythos was reportedly capable of infiltrating almost all classified government networks within hours — underscoring the rapid pace at which AI capabilities can outstrip existing safeguards.
The incident raises urgent questions about how AI should be regulated. While there is widespread agreement that oversight is necessary, the current approach by the Trump administration has been characterized by critics as inconsistent and driven by ad hoc decisions. Only weeks before the Anthropic order, President Donald Trump signed an executive order establishing a voluntary review framework for frontier AI technologies, aiming to develop clear protocols within 60 days. However, before these procedures were in place, the Pentagon labeled Anthropic a “supply-chain risk,” and several federal agencies were instructed to halt use of its systems.
Government concerns partly stem from Anthropic’s handling of access permissions for Mythos, with reports suggesting the company exceeded authorized distribution and delayed responses to official inquiries. Intelligence officials reportedly advocated for a stringent government reaction. Still, experts caution that the lack of predictable, institutionalized processes risks undermining trust and stability in the rapidly evolving AI sector.
The administration’s approach has drawn criticism for its perceived reliance on raw authority rather than transparent rules. Observers note a pattern of targeted government interventions in various industries, including investments and tax actions, sometimes linked to political considerations rather than consistent policy. Defense Secretary Pete Hegseth publicly accused Anthropic of “betrayal” and “duplicity,” while some senior officials criticized the company on ideological grounds. President Trump also mocked the company publicly, reflecting strained relations.
Experts argue that the significance of AI demands a governance model akin to that of the U.S. Federal Reserve for banking. Such an institution would combine public oversight with expert knowledge, enforce clear, comprehensive standards, and implement graduated enforcement ranging from warnings to restrictions — rather than unpredictable bans. This framework would apply uniformly to all developers, fostering transparency and stability.
The call is for the United States and other democracies to establish independent regulatory bodies to manage the risks and rewards of AI, coordinating internationally much as central banks coordinate financial systems. Advocates assert that regulation and innovation are not mutually exclusive and that well-regulated sectors tend to thrive sustainably.
With AI poised to become a foundational technology shaping business, government, and daily life, the incident involving Anthropic spotlights the urgent need for a coherent, expert-driven governance system that balances rapid technological advancement with national security, economic stability, and public trust.