Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell took the extraordinary step this week of summoning the CEOs of America's largest banks to an urgent meeting at Treasury headquarters in Washington, warning them about cybersecurity risks posed by Anthropic's powerful new Claude Mythos AI model.
The meeting, which took place on Tuesday, underscores growing alarm among top U.S. regulators that frontier AI models are fundamentally changing the cybersecurity threat landscape — and that the financial sector must act quickly to shore up its defenses.
Who Was in the Room
The gathering brought together some of the most powerful figures in American finance. Attendees included Citigroup CEO Jane Fraser, Morgan Stanley CEO Ted Pick, Bank of America CEO Brian Moynihan, Wells Fargo CEO Charlie Scharf, and Goldman Sachs CEO David Solomon. JPMorgan's Jamie Dimon was reportedly the only major banking CEO unable to attend.
Why Mythos Has Washington's Attention
The meeting was prompted by Anthropic's own disclosures about Claude Mythos Preview, a general-purpose AI model that has demonstrated striking cybersecurity capabilities. According to Anthropic's published assessment, Mythos has identified thousands of previously unknown zero-day vulnerabilities across every major operating system and web browser during its initial testing phase.
Among its discoveries: a 27-year-old vulnerability in OpenBSD — an operating system widely regarded as one of the most secure — and a 16-year-old flaw in FFmpeg, a ubiquitous video software library that automated testing tools had failed to catch despite running the affected code millions of times.
Anthropic has stated that Mythos was not specifically trained for cybersecurity; its capabilities emerged from its strong coding and reasoning abilities. The company concluded that unrestricted public access to the model poses unacceptable risks of misuse.
Project Glasswing: Controlled Deployment
Rather than releasing Mythos publicly, Anthropic launched Project Glasswing earlier this week, a cybersecurity initiative that provides controlled access to the model through a coalition of 11 major technology and finance companies. Launch partners include Amazon, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.
Anthropic is committing up to $100 million in usage credits for Claude Mythos Preview across the initiative, along with $4 million in direct donations to open-source security organizations including the Linux Foundation's Alpha-Omega and OpenSSF programs and the Apache Software Foundation.
Implications for the Financial Sector
The Bessent-Powell meeting signals that U.S. regulators view AI-powered cyber threats as a systemic risk to financial stability — not just an IT concern. By bringing bank CEOs directly into the conversation, the administration is pushing for board-level awareness and preparedness.
For the broader AI industry, the episode highlights a growing tension: models capable enough to find and fix critical vulnerabilities are also capable enough to exploit them. How companies, regulators, and governments navigate that duality will likely define cybersecurity policy for years to come.
