Federal Reserve Vice Chair for Supervision Michelle Bowman addressed the dual nature of artificial intelligence in banking at a Financial Stability Oversight Council roundtable on May 1. She identified Anthropic’s Mythos model as a tool that can both identify infrastructure vulnerabilities and be exploited by malicious actors if left unsupervised.
In April 2026, Treasury Secretary Scott Bessent and Fed Chair Jerome Powell held meetings with the CEOs of JPMorgan Chase, Citigroup, Bank of America, and Goldman Sachs. The discussions focused on cybersecurity risks associated with the Mythos model and the rapid advancement of AI capabilities that currently outpace existing regulatory frameworks.
Bowman highlighted the gap between large and small financial institutions, noting that smaller banks often lack the resources to develop independent AI defenses. The Federal Reserve is currently working with the Federal Financial Institutions Examination Council and payment networks like Mastercard and Visa to address these systemic security disparities.
The regulator acknowledged that current model risk management guidelines were established before the mainstream adoption of generative and agentic AI. While the Fed is evaluating how to oversee autonomous AI agents capable of executing transactions, no specific new rules have been proposed yet.
