OCC examiners may use the November 2026 AI framework (Bulletin 2026-54) when supervising community banks, but they are told to tailor exams to the bank's size, complexity and risk profile and to how consequential its AI uses are. The framework sets no new expectations for banks. It covers only OCC-supervised banks, and for most small banks the main route to AI is a vendor, which section 2.9 addresses.
TRACKED DOCUMENTS: OCC Bulletin 2026-54 · OCC Bulletin 2023-17 · OCC Bulletin 2026-13
Does the OCC AI framework apply to community banks?
Yes, as a tool for examiners. Bulletin 2026-54 says: “Examiners may use the framework for supervising community banks’ SI use.”
The OCC calls artificial intelligence “Super Intelligence (SI)”. This page writes AI, except inside quotations.
The framework is written for OCC examiners, not for banks. The bulletin says it “does not establish supervisory expectations for banks” and that banks “are not expected to use this document to assess their use of SI or their SI risk management”. It has no effective date and no deadline. Put simply, it tells you what an examiner may look at when your bank uses AI. It does not give your bank new rules. For the whole picture of what the OCC has issued, start with the plain-English overview.
Is my bank supervised by the OCC at all?
Only if it is a national bank or a federal savings association (or a federal branch or agency of a foreign bank). That is how footnote 1 of the bulletin defines “banks”.
Many community banks hold a state charter. A state-chartered bank is supervised by its state banking department and by either the Federal Reserve or the FDIC. The OCC framework does not speak for those agencies. A state bank can still read it as a picture of how one federal supervisor thinks about AI, but it is not their examiners' document. In practice, ask your compliance officer which agency examines your bank.
What does proportionate mean for a small bank?
It means examiners are told to scale their work to the bank and to how risky each AI use is. A small bank with a few low-risk tools need not get the same review as a large bank running AI agents.
The OCC says this in several places:
- The bulletin's community-bank note: examiners “should tailor their examinations to the bank’s size, complexity, and risk profile, proportionate to the nature, consequences, autonomy, and complexity of the bank's SI use cases.”
- Section 1: examiners “should apply judgement, in conjunction with consideration of the bank’s size, complexity, and risk profile.”
- Section 2.2: a consequence-based assessment may help a bank apply more intensive controls to consequential uses “without imposing the same control structure on lower-risk activities.”
- Section 2.13: “Lower materiality uses may warrant limited testing.” Section 2.14 says documentation “may be proportionate to the use case”. Section 2.15 says audit coverage may be scaled to the bank's AI use cases.
- Section 2.4: board members and managers may need technical knowledge “proportionate to the bank's use cases.”
Section 1.1 adds the key idea. Risk depends on what the AI can do, not on how advanced it is. Its own example: “a simple rules-based agent authorized to initiate transactions could present greater risk than a sophisticated model limited to internal drafting assistance.” Put simply, a small bank's exposure grows with what it lets the tool touch and decide.
What kinds of AI use does the OCC put at each risk tier?
Section 3.2 gives four illustrative tiers, from Tier 4 (Low) to Tier 1 (Critical), each with example uses. The examples are the OCC's. Where your bank's tools fall is a judgment for the bank and its examiners, not something this page can decide.
| Tier (section 3.2) | Profile in the text | The OCC's example uses |
|---|---|---|
| Tier 4: Low | No customer or financial data exposure, no direct action, fully human-controlled use | Internal productivity tools; code assistance; general employee Q&A |
| Tier 3: Moderate | Decision support with mandatory human-in-the-loop, limited data sensitivity, reversible | Internal document summarization for credit memos; agent-assisted research |
| Tier 2: High | Direct action or high-weight decision support, human-on-the-loop, moderate-to-high impact | AI-assisted underwriting; agent-initiated customer communications; AML alert disposition; code assistance for internal code or data access changes |
| Tier 1: Critical | Direct action, low or no reversibility, high financial or customer impact, autonomous or partially autonomous | Autonomous payment agent; real-time credit decisioning; fraud-hold release; market access |
Two terms help. Human-in-the-loop means a person approves each output or action before it is used. Human-on-the-loop means a person watches and can step in (section 2.3). AML alert disposition means deciding what to do with an anti-money-laundering alert, the kind of work a BSA (Bank Secrecy Act) officer does.
Section 3.1 gives the anchor cases: “a simple internal productivity tool with no customer or financial data exposure sits at Tier 4, whereas a fully autonomous, customer-facing payment agent sits at Tier 1.” For example, many everyday small-bank uses look more like the Tier 4 and Tier 3 examples than the Tier 1 ones. That is a resemblance, not a ruling. More on the method in the four OCC AI risk tiers.
Why does vendor AI matter most for a community bank?
Because most small banks buy AI rather than build it, and section 2.9 is about exactly that. It opens: “Banks frequently obtain SI capabilities from third parties rather than develop them internally.”
A third party is any outside company the bank relies on, such as a core processor or software vendor. Section 2.9 lists points examiners may consider where a provider's AI can affect the bank's decisions, customers or data:
- Change and exit. Existing vendors may add AI features to products the bank already uses. The bank may require notice and may retest or reassess.
- Liability. How contracts assign responsibility for model-related errors or losses, including customer harm.
- AI incidents. The bank may define terms and notification timelines for events such as hallucination (fluent but false output), misfire, drift or manipulation.
- Provider agent visibility. Limits on what a provider's AI agents may do inside bank systems and data.
Section 2.4 adds that even when vendors build or host the system, “accountability for managing the bank's risk generally remains with the bank.” Section 2.7 notes contracts may say whether bank data can train models offered to other customers.
The wider third-party guidance applies to small banks too. The community-bank note in Bulletin 2023-17 says: “This guidance applies to all banks with third-party relationships.” Section 2.9 says AI vendor review may supplement that guidance. See vendor and third-party AI for the full list.
Does the 2026 model risk guidance cover a community bank's AI?
Mostly not for generative AI. Bulletin 2026-13 says generative and agentic AI models are outside its scope, and that it is expected to be most relevant to banks with over $30 billion in total assets.
Model risk is the chance of harm from decisions based on a wrong or misused model. Bulletin 2026-13's community-bank note reads: “This guidance is applicable to all community banks, subject to the limitations discussed in the guidance.” The bulletin also says it may be relevant to smaller banks with significant model risk or activities outside traditional community banking. Footnote 1 of the framework says the 2026-13 principles apply to non-generative and non-agentic AI models, so the framework is more applicable to generative and agentic AI. For example, a vendor credit scorecard built on machine learning sits closer to 2026-13, and a chat assistant sits closer to the framework. Details in model risk management for generative AI.
Where can a small bank with no AI team start?
With a list of the AI you already use and what each tool is allowed to do. The steps below are our suggestions, not OCC requirements. Each points to the section that describes the idea.
- List every AI tool in use, including AI features your vendors have switched on in existing products (sections 2.1 and 2.9).
- For each, write one paragraph: what it does, what it may not do, who may use it, and what data it can reach (section 2.1).
- Note whether it only helps a person or acts on its own, and whether a mistake can be undone (sections 2.2 and 3.1).
- Name one owner per tool and decide who approves new uses (section 2.4).
- Ask each vendor about notice of AI changes, data use for training, and who pays for errors (sections 2.7 and 2.9).
- For any tool that drafts text for customers, decide who checks it before it goes out (sections 2.3 and 2.8).
- Give the board a short yearly summary of AI use and incidents (sections 2.4 and 2.16).
For what an examiner may ask to see, read preparing for an OCC exam of your AI. For board roles, see board, management, audit and reporting.
Which OCC tier examples do common small-bank AI uses resemble?
The left column lists our own examples of common uses. The middle column names the section 3.2 example each most resembles. This is a reading aid, not a tier assignment. The real answer depends on data access, autonomy and reversibility (section 3.1).
| Use (our example) | Closest OCC example (section 3.2) | Passage to read |
|---|---|---|
| Staff drafting internal emails or policies with a chat assistant, no customer data | Internal productivity tools (Tier 4) | Sections 3.1, 2.7 |
| Staff asking an assistant how a procedure works | General employee Q&A (Tier 4) | Sections 2.1, 2.13 |
| Summarizing financial statements for a loan officer's memo | Internal document summarization for credit memos (Tier 3) | Sections 2.3, 2.13 |
| Chatbot on the website that answers customers by itself | Agent-initiated customer communications (Tier 2) | Sections 2.8, 2.13 |
| Vendor tool that suggests closing AML alerts | AML alert disposition (Tier 2) | Sections 2.3, 2.8 |
| Vendor fraud tool that releases holds without staff review | Fraud-hold release (Tier 1) | Sections 2.10, 3.1 |
For depth on any section, use the section-by-section reference or the OCC AI regulation tracker.
Does my community bank have to follow the OCC AI framework?
The bulletin says the framework does not establish supervisory expectations for banks and sets no enforceable standards. It is a resource for examiners, who may use it when they look at a community bank's AI. It has no effective date or deadline.
We are a state-chartered bank. Does this apply to us?
The framework is written for OCC examiners, who supervise national banks and federal savings associations. A state bank is examined by its state and by the Federal Reserve or the FDIC, and the OCC document does not speak for them. It can still be useful reading.
We only use AI built into our vendors' products. Does that count?
Yes. Section 2.9 starts from the fact that banks often get AI from third parties, and section 2.4 says accountability generally stays with the bank. A vendor switching on a new AI feature is one of the changes section 2.9 mentions.
Will examiners treat a small bank like a large one?
The bulletin says examiners should tailor exams to the bank's size, complexity and risk profile, proportionate to its AI use cases. Section 2.2 also describes not imposing the same controls on lower-risk uses.
Do we need a separate AI risk appetite?
No. The framework says banks are not required to define a separate AI risk appetite and that examiners should not treat AI as a separate risk category. AI risk is handled inside existing categories such as operational and compliance risk.
Does Bulletin 2026-13 make us validate our chatbot?
Bulletin 2026-13 says generative and agentic AI are outside its scope and that it is most relevant to banks over $30 billion in assets. The framework, not 2026-13, is the OCC text more applicable to a chatbot.
Is a staff chat assistant low risk?
Section 3.2 lists internal productivity tools and general employee Q&A as Tier 4 examples, with no customer or financial data exposure. If the tool can reach customer data or act on its own, the factors in section 3.1 point higher.
- OCC Bulletin 2026-54: Super Intelligence: Supervisory Framework — Office of the Comptroller of the Currency, Nov 4, 2026 (occ.gov) · our page on OCC Bulletin 2026-54
- Super Intelligence: Supervisory Framework (full text, PDF) — Office of the Comptroller of the Currency, Nov 4, 2026 (occ.gov) · our page on OCC Bulletin 2026-54
- OCC Bulletin 2026-13: Model Risk Management: Revised Guidance — Office of the Comptroller of the Currency, Apr 17, 2026 (occ.gov) · our page on OCC Bulletin 2026-13
- OCC Bulletin 2023-17: Third-Party Relationships: Interagency Guidance on Risk Management — Office of the Comptroller of the Currency, Jun 6, 2023 (occ.gov) · our page on OCC Bulletin 2023-17
When this rule moves, the next morning's brief says so
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning