U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell recently convened an urgent meeting with bank chief executives, cautioning them about significant cyber risks posed by Anthropic’s latest artificial intelligence model. Anthropic, an artificial intelligence startup known for developing advanced AI models, has introduced its powerful Mythos model, which has prompted high-level discussions regarding its potential impact on financial security. The company chose not to proceed with a broad release of Mythos, citing concerns it could expose previously unknown cybersecurity vulnerabilities across critical systems.
The AI developer stated that its Mythos model is capable of identifying and exploiting weaknesses across “every major operating system and every major web browser.” Anthropic had previously indicated it was engaged in ongoing discussions with U.S. government officials regarding the model’s “offensive and defensive cyber capabilities.” A source close to the matter confirmed Anthropic’s proactive outreach, noting the company briefed senior U.S. government officials and key industry stakeholders on Mythos’s capabilities well in advance of its release. Access to Mythos will be carefully controlled, limited to approximately 40 technology companies, including industry giants like Microsoft and Google.
The Treasury-hosted meeting in Washington on Tuesday was specifically aimed at ensuring banks are aware of the risks posed by Mythos and similar models, and are actively taking steps to defend their systems. Invitations for the critical discussion were extended whilst most CEOs of the largest U.S. banks were already in the capital for other engagements. Reportedly, the chief executives of Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs were present. However, JPMorgan CEO Jamie Dimon was unable to attend. Goldman Sachs, Wells Fargo, and the Federal Reserve declined to comment on the proceedings, whilst the Treasury, other lenders, and Anthropic did not immediately respond to requests for comment.
