AI in action · Regulation

OCC Tells Examiners to Judge Bank AI by What Could Go Wrong

The OCC's November 4, 2026 framework tells bank examiners to scale AI reviews to the consequences of failure, not to how advanced the tool is. It adds no new rules for banks but shows what examiners will ask.

The Office of the Comptroller of the Currency, which supervises US national banks and federal savings associations, published Bulletin 2026-54, "Super Intelligence: Supervisory Framework," on November 4, 2026. Signed by James M. Gallagher, Senior Deputy Comptroller and Chief National Bank Examiner, it is written for the OCC's own examiners. The agency is explicit that the document "does not establish supervisory expectations for banks" and creates no new enforceable standards.

The vocabulary follows the administration's. Under Executive Order 14434, federal agencies now call artificial intelligence "super intelligence," or SI, and the bulletin's footnote defines it as the technologies previously described as AI under federal law, together with their fast-advancing capabilities. The practical content, though, is plain supervision: examiners are told to "tailor their examination of banks' use of SI based on the materiality of the SI use case."

The central idea is that risk comes from consequences, not sophistication. The OCC writes that "control intensity could increase with the potential consequences of an SI agent's failure or misuse," and that a bank's classification of each use case should reflect potential consequences, how much authority is handed to the system and how much human oversight remains. A simple chatbot answering branch-hours questions and an autonomous agent that moves customer money are different problems, even if the second uses a less advanced model.

Examiners are also told "not to treat SI as a separate and distinct risk category." Instead, AI considerations are layered onto existing model-risk management, governance, technology, compliance and operational-risk work, and banks do not need a separate AI risk appetite. That fills a gap the agencies left in April, when Bulletin 2026-13 revised model-risk guidance but placed generative and agent-based systems outside its scope. Community banks are covered too, with reviews scaled to their size and complexity.

For bank investors, the framework lowers one kind of uncertainty. It signals that the OCC does not plan a separate AI rulebook, so the cost of deploying AI should sit inside existing compliance budgets rather than a new one. It also tells banks where scrutiny will concentrate: customer-facing agents that can act on accounts, and contracts with outside AI vendors, since the bank stays responsible for systems it buys. The OCC says it will update the guide as the technology and its own understanding develop.

Sources: OCC — Bulletin 2026-54, Super Intelligence: Supervisory Framework (Nov 4, 2026); Reed Smith — OCC issues supervisory framework on use of AI (Nov 2026)

By Trade2Spike Team, in-house editorial team. About us · Methodology