Washington Summons Bank Chiefs Over New AI Security Model

The heads of America's largest banks convened for an unplanned meeting in Washington this week, called by Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent. The urgent topic: the potential cybersecurity implications of Anthropic's newly released Claude Mythos Preview AI model. CNBC verified the gathering on Friday.

Sources indicate the bank CEOs, already in the capital for a scheduled Financial Services Forum meeting, were pulled into a separate session to examine risks that the advanced AI might be weaponized by malicious actors. This concern prompted Anthropic to release Mythos in a tightly controlled preview.

The model is part of a security-focused effort called Project Glasswing. JPMorgan Chase is a named partner in the initiative, alongside tech heavyweights Apple, Google, Microsoft, and Nvidia.

Anthropic stated it has maintained consistent dialogue with U.S. government agencies, specifically naming the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation, regarding the system's capabilities. The company did not provide further comment on the high-level bank meeting when contacted.

Reports of the confidential discussion were initially published by Bloomberg and the Financial Times.

Source: CNBC