The Office of the Comptroller of the Currency and the Federal Reserve have begun incorporating detailed AI-related inquiries into routine bank examinations. Supervisors are targeting how institutions manage third-party vendors, protect sensitive client data, and maintain operational control. Banks are now being pressed to demonstrate the presence of "kill switches" and clear human oversight frameworks to prevent autonomous systems from accessing unauthorized data or overstepping governance guardrails.

Despite the pressure, agencies are not yet drafting new, AI-specific regulations. Instead, they are applying existing frameworks governing model risk management and consumer protection to the new technology. This approach allows regulators to gather intelligence on how firms are grappling with rapidly evolving models, such as Anthropic’s frontier systems, without the constraints of rigid rulemaking that might be rendered obsolete by the industry's pace of innovation.

The challenge remains the velocity of development. As Federal Reserve Vice Chair for Supervision Michelle Bowman noted in May, while current supervisory tools are designed to encourage sound governance, agencies must continuously evaluate if their guidance remains fit for the future. For now, the focus is on forcing banks to prove they have contingency plans for system failures and that their third-party providers adhere to the same stringent security standards required of the financial institutions themselves.