Fed & Treasury Warn Banks on AI Cyber Risk: Mythos Model Threat
Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened an urgent, unprecedented meeting with Wall Street leaders to address the systemic cyber risks posed by advanced AI models, specifically citing capabilities demonstrated by Anthropic's “Mythos” model. This immediate action underscored the officials' belief that the financial system's existing cyber defenses are insufficient against rapidly evolving AI threats.
The core concern revolves around Mythos's reported ability to discover thousands of high-severity vulnerabilities, including zero-days, across every major operating system and web browser, with over 99% still unpatched. This capability dramatically compresses the timeline between vulnerability discovery and weaponized exploitation, presenting an acute risk to the interconnected financial sector which relies heavily on shared cloud providers, software vendors, and critical infrastructure. The officials' urgency was informed by Anthropic's own advance briefings on Mythos, despite the government simultaneously terminating its contracts with Anthropic due to national security and procurement concerns.
To mitigate immediate risks, Anthropic launched “Project Glasswing,” restricting Mythos access to a consortium of launch partners, including major tech and financial firms like JPMorgan, and committing resources to open-source security. This initiative represents a “bull case” scenario where controlled access allows vulnerabilities to be patched before widespread exploitation. However, the “bear case” anticipates more models with similar offensive capabilities, potentially leading to stricter supervisory expectations from regulators, including mandatory vendor concentration reviews and tighter incident reporting.
This urgent meeting operationalized a risk framework Washington had been developing for years, recognizing AI's potential to create systemic vulnerabilities through third-party dependencies and cyberattacks. The move signals a shift from theoretical discussions to active preparation, acknowledging that the balance between AI offense and defense is narrowing rapidly, necessitating proactive measures to safeguard financial stability.
