U.S. State Regulators Unveil AI Examination Framework

The Federal Reserve, Office of the Comptroller of the Currency and Federal Deposit Insurance Corp. updated joint model-risk guidance in April 2026 but left generative and agentic AI outside its scope, citing the technologies’ novelty and rapid evolution. That gap matters because such systems can access sensitive customer data, influence lending or pricing decisions and, in the case of AI agents, take actions with limited human direction, broadening governance, operational and consumer-protection risks.

The Conference of State Bank Supervisors on Sept. 16, 2026, released an AI Supervisory Framework covering state-chartered banks and state-licensed nonbanks. Its five documents include a core examiner guide, a 28-page work program, nonbank supplements, a risk-tiering worksheet and source materials. State regulators oversee 3,355 of 4,233 FDIC-insured banks and thrifts, or 79%, meaning most could face AI questions at their next examination. Adoption is discretionary for each state and creates no new legal obligations.