The US Office of the Comptroller of the Currency has issued a framework to help examiners assess banks’ use of AI, including generative and agentic systems. It is guidance for supervision, not a new set of enforceable bank rules, according to PYMNTS’ account of the OCC’s announcement.
Watch Desk analysis
What happened
The OCC says examiners may use the framework to identify AI-specific considerations and tailor supervisory work to a bank’s use cases and their potential consequences. Those considerations supplement existing practices for model risk, governance, technology, compliance and operational risk.
The framework does not make AI a separate risk category, require banks to define a separate AI risk appetite, or set prescriptive standards. The OCC describes its approach as technology-neutral: banks should conduct their activities safely and soundly whether they use traditional methods or newer technologies.
Why it matters
Banks are exploring generative AI and agents, and examiners now have a framework for considering those uses within existing oversight. That makes the document relevant to how banks and supervisors discuss AI risk, without turning it into a checklist of new obligations for every bank.
The distinction is worth keeping clear. A supervisory framework can shape what examiners look at, but the OCC’s account says it does not itself establish new expectations or enforceable requirements for banks.
Our read
This is a signal about how the regulator intends to examine AI use, not a rulebook banks must now implement line by line. The useful question is whether applying existing risk practices to AI gives examiners enough to work with, or whether specific uses will eventually need clearer standards.
What to watch
- How OCC examiners apply the framework to specific bank AI use cases.
- Whether the OCC later issues more detailed expectations or requirements.
- How banks govern agentic AI and keep appropriate human oversight in place.
Discussion spark: Should bank supervisors assess AI through existing risk frameworks, or do generative and agentic systems need their own explicit standards?
Sources and evidence
- Source update (4 November 2026, 15:51 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.