Bessent and Powell Warn Bank CEOs on Anthropic Mythos Risks

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened a meeting at the Treasury Department in Washington on Tuesday to warn bank chief executives that Anthropic’s new AI model, Mythos, can expose previously unknown cybersecurity vulnerabilities.

Anthropic launched Mythos earlier this week but has not issued a broad release, citing concerns that the model could reveal new attack paths. The company has described Mythos as able to identify and exploit weaknesses across major operating systems and web browsers. Access to the model is limited to about 40 technology firms, including Microsoft and Google, and Anthropic has been briefing U.S. officials on the model’s capabilities.

Invitations for the Treasury-hosted session went out while many top bank leaders were already in Washington for other meetings, allowing a rapid assembly of industry and regulators. Executives from Citigroup, Morgan Stanley, Bank of America, Wells Fargo and Goldman Sachs attended. JPMorgan Chase CEO Jamie Dimon was unable to join.

Officials urged banks to examine and strengthen cyber defenses in light of Mythos’s capabilities. The discussion covered network monitoring, patching practices, incident response readiness and channels for coordination between firms and government agencies if new vulnerabilities are discovered or exploited.

Banking sources described the briefing as focused on operational risk management rather than new regulatory measures. Anthropic provided briefings to senior U.S. officials and select industry stakeholders before limiting access to the model, and has continued discussions with federal authorities about both offensive and defensive cyber implications.

Goldman Sachs and the Federal Reserve declined to comment. Treasury officials, the banks present and Anthropic did not immediately provide additional comment.

Security researchers have noted that AI models able to analyze code and system behavior can help identify novel exploits. Mythos’s limited launch has prompted attention to disclosure practices, access limits and coordination between the private sector and government when tools reveal potential vulnerabilities.