US and UK Regulators Review Anthropic's Mythos AI Model

Financial regulators in the United Kingdom and the United States are conducting urgent risk reviews of Anthropic's new AI model, Claude Mythos. The Bank of England and the Financial Conduct Authority are coordinating with the National Cyber Security Centre to determine if the model threatens critical financial IT systems. In the U.S., Treasury Secretary Scott Bessent met with major Wall Street banks, and administration officials have encouraged firms like JPMorgan Chase and Goldman Sachs to test the model.

Anthropic is deploying the model under Project Glasswing for defensive cybersecurity, and the U.S. government has authorized limited access for select American cybersecurity firms. However, the company remains embroiled in a legal conflict with the Department of Defense. The Pentagon labeled Anthropic a supply-chain risk and barred it from military business after CEO Dario Amodei refused to allow the AI to be used for autonomous weapons or mass surveillance.

While President Donald Trump ordered federal agencies to cease using Anthropic's Claude model, some agencies are quietly sidestepping the ban. The Department of Commerce is currently testing Mythos' hacking and cyber-scanning capabilities, and several congressional committees have requested briefings. Co-founder Jack Clark confirmed that Anthropic continues to brief the Trump administration on the model's capabilities, arguing that the government must be informed about such technology to protect national security despite the ongoing contracting dispute.