Federal Reserve Warns Banks on AI Cybersecurity Risks

Federal Reserve Vice Chair for Supervision Michelle Bowman addressed banks on Thursday regarding the integration of artificial intelligence into financial systems. Speaking at a Financial Stability Oversight Council roundtable on May 1, Bowman described AI as both a powerful tool for defending against cyberattacks and a significant threat when utilized by malicious actors.

Bowman identified Anthropic’s Mythos model as a primary example of this duality. While the model can identify vulnerabilities within banking infrastructure, she noted that it could also be repurposed to exploit those same systems if not properly supervised.

The remarks follow a series of meetings in April 2026 between Treasury Secretary Scott Bessent, Fed Chair Jerome Powell, and the CEOs of JPMorgan Chase, Citigroup, Bank of America, and Goldman Sachs. These discussions focused on the cybersecurity risks posed by Mythos and the broader implications of rapidly advancing AI capabilities for the financial sector.

JPMorgan CEO Jamie Dimon has long identified cybersecurity as a primary risk for the bank, describing the current environment as a transitory period where AI capabilities are outpacing the industry's defensive measures.

Bowman acknowledged that existing regulatory frameworks, including the Fed’s model risk management guidance, were established before the mainstream adoption of generative and agentic AI. She noted that these older guidelines were designed for traditional statistical models rather than systems capable of generating code or autonomously executing workflows.

A significant concern raised by Bowman is the resource gap between large and small financial institutions. She suggested that the Federal Financial Institutions Examination Council could provide support to smaller banks that lack the scale to build advanced AI defenses independently. The Fed is also coordinating with payment networks like Mastercard and Visa to address potential security gaps across the financial system.

While Bowman indicated that the Federal Reserve is working to address these regulatory challenges, she did not propose specific new rules. The direct engagement between top policymakers and bank executives regarding specific AI models underscores the growing systemic concern surrounding frontier AI technology.