Confirmed 4 Nov 2026 · Governance A one-minute read
America's bank regulator told examiners to judge AI by consequences, not sophistication.
On November 4 the Office of the Comptroller of the Currency issued its first AI supervisory framework, Bulletin 2026-54. Caveat: it sets no new requirements for banks.
Bank supervisors would rate AI risk by how advanced the model is.
They rate it by what breaks, who acts, and what cannot be undone.
Why it matters
Examiners decide what banks actually fix. This framework tells them a simple agent moving money matters more than a clever model writing memos. It also puts the government's term, Super Intelligence, into a banking rulebook for the first time, which binds bank supervision to the administration's AI vocabulary.
Established
- The OCC issued Bulletin 2026-54 on November 4, titled Super Intelligence: Supervisory Framework.
- It tells examiners not to treat AI as a separate risk category.
- Risk rests on consequences, transfer of agency and human oversight, not sophistication.
Not established
- That banks must do anything. The framework sets no supervisory expectations.
- The risk tiers and shutdown controls. Those are reported from the attached guide.
- That any bank has yet been examined against the framework.
The five-minute readThe record, the timeline, the weight and 4 sources
The record
Verified: we read Bulletin 2026-54 at occ.gov, including its title, date, audience and quoted sentences. It surfaced late. The bulletin is dated November 4 and Editions 074 and 075 covered that window without reaching it, so we card it under its own event date. The risk tiers and shutdown language sit in the attached guide, which would not open, and are marked reported.
The Office of the Comptroller of the Currency, which supervises every national bank in the United States, issued Bulletin 2026-54 on Wednesday November 4, 2026, titled "Super Intelligence: Supervisory Framework", release identifier OCC 2026-54. We read it at occ.gov. It is addressed to chief executive officers of all national banks, federal savings associations and federal branches and agencies, to department and division heads, and to all examining personnel. The framework is "designed to support OCC examiners in evaluating banks' use of super intelligence (SI)" in normal-course business activities. On method, examiners "should not treat SI as a separate and distinct risk category", and risk classification "may be based on potential consequences, transfer of agency" and the degree of human oversight, "rather than the technical sophistication of the SI used". On weight, "control intensity could increase with the potential consequences" of an "SI agent's failure or misuse". On force, it "does not establish supervisory expectations for banks", and "nothing in this guide is intended to establish new regulatory expectations". The bulletin links a guide of the same title, which returned a server error every time. Reported, from Decrypt, Forbes and Forvis Mazars, which agree: the guide runs 22 pages and sets 4 illustrative risk tiers, with fully autonomous payment agents and real-time credit decision systems highest and basic internal productivity tools lowest; it tells examiners to look at identity verification, delegated permissions, transaction limits, monitoring and the ability to suspend or revoke access; and it raises external agents run by customers, financial technology firms or other businesses that connect through bank channels. The same accounts report that it suggests circuit breakers, rate limits, restricted permissions and emergency shutdown controls for systems that keep operating while making harmful decisions, and that a press release carries identifier NR-OCC-2026-97A. Context from Axios on October 22, as reported: Morgan Stanley was blocking outside agents while JPMorgan, Bank of America and Citi restricted access to theirs. Not established: the tier definitions in the agency's own words, whether any bank has been examined against the framework, and whether the other federal banking regulators will follow.
How it got here
- Oct 2
The executive order Inaugurating the Era of Super Intelligence is published in the Federal Register, introducing the term the OCC has now adopted.
- Oct 22
Axios reports, as relayed by Decrypt, that Morgan Stanley is blocking outside AI agents from customer accounts and that JPMorgan, Bank of America and Citi are restricting their access.
- Nov 2
Hong Kong's central bank opens a project to govern AI agents that pay, carded in Edition 070. The question of who an agent acts for is now live in 2 financial centres.
- Nov 4
The OCC issues Bulletin 2026-54, Super Intelligence: Supervisory Framework, addressed to every national bank chief executive and to all examining personnel, creating no new requirements.
- Nov 5
The Education Department proposes striking the words artificial intelligence from its own priorities, carded in Edition 075. A second agency moves to the new vocabulary.
Every national bank and federal savings association, whose next examination now has a script; the Federal Reserve and the Federal Deposit Insurance Corporation, which supervise the banks the OCC does not; Morgan Stanley, JPMorgan, Bank of America and Citi, which already blocked outside agents; the agent companies selling into banks; and the bank technology vendors the guide reportedly reaches.
What to watch
- Would strengthen itThe OCC, the Federal Reserve or the Federal Deposit Insurance Corporation issues guidance, an enforcement action or a rule that addresses customer-directed or third-party AI agents by name, by March 31, 2027.
- Would weaken itNo federal banking regulator takes any further action on AI agents, and no examination finding citing Bulletin 2026-54 becomes public, by September 30, 2027.
- Would strengthen itA third federal agency issues a rule, bulletin or priority using Super Intelligence in place of artificial intelligence in its title, by June 30, 2027.
Why it weighs this much
Cost is low: writing guidance costs a regulator little, and the bulletin explicitly imposes nothing on banks. Commitment is medium, because a numbered bulletin addressed to every national bank chief executive and to all examining personnel is a public position the agency must now examine against, even without new rules. Reversibility is easy: non-binding guidance can be withdrawn or rewritten by the next comptroller without notice or comment. Persistence is years, because examiner handbooks outlive the news cycle, and because the framework's method, consequence and agency rather than sophistication, is the kind of test other regulators copy.
The words
- Super Intelligence (SI)
- The term this administration now uses for AI in federal documents. An October order introduced it, and the bank regulator has now used it in a supervisory title.
- Transfer of agency
- How much authority to act a person hands to software. The more a system decides and executes alone, the more agency has moved.
- Supervisory framework
- Guidance written for examiners, not for banks. It changes no rule, but it shapes what examiners ask and therefore what banks fix.
Sources
- OOCC Bulletin 2026-54, Super Intelligence: Supervisory Framework, issued Nov 4, 2026, read at occ.govocc.gov
- DDecrypt: US bank regulator issues first AI framework as banks battle rogue agents, Nov 4, 2026decrypt.co
- FForbes: OCC gives examiners a playbook for bank use of Super Intelligence, Nov 5, 2026forbes.com
- FForvis Mazars: OCC AI framework, consequences not sophistication drive scrutiny, Nov 2026forvismazars.us
Your call
Which explanation do you believe?
A call is final. Other people's calls show after yours.