In a recent development, Fed Chair Jerome Powell and Treasury Secretary Scott Bessent urgently gathered Wall Street executives for a critical discussion on AI-driven cyber threats. This move bypassed usual briefing schedules to address the risks associated with models like Mythos directly with bank CEOs.
The meeting underscored the systemic nature of these risks as emphasized by the high-level involvement. Meanwhile, Treasury and other agencies were concurrently managing issues with Anthropic’s technology, highlighting a complex situation involving both national security and financial stability.
Anthropic claims that its Mythos model has identified thousands of significant vulnerabilities in major operating systems and web browsers, with over 99% remaining unpatched. The company responded by restricting access through Project Glasswing, engaging partners such as Amazon Web Services, Apple, and Microsoft among others to manage these risks responsibly.
These developments are crucial for banks that rely heavily on shared software ecosystems. The Treasury’s Financial Services Sector Risk Management Plan highlights cloud concentration and AI as significant sector risks due to the potential cascading failures from common vulnerabilities.
Policy efforts have been underway since 2024, with initiatives like the AI Innovation Series aiming to bolster financial resilience against these emerging threats. Treasury’s recent actions reflect a proactive stance in operationalizing long-term risk frameworks.
Officials are now warning of the urgent need to address these risks, acknowledging that Mythos-class capabilities pose substantial operational challenges. The future may see either continued coordinated efforts to patch vulnerabilities or stricter regulatory measures if new models with similar offensive capabilities emerge.
This meeting marks a clear acknowledgment by Powell and Bessent that the pace at which these threats evolve necessitates a faster adaptation of the financial system’s cybersecurity posture.