Summary

  • A banking organization supervised by the OCC must notify its federal regulator about a "notification incident" as soon as possible and no later than 36 hours after the bank determines that such an incident has occurred.
  • The same rule times bank service providers with a different instrument: covered services degraded "for four or more hours" require notice "as soon as possible" — a threshold with no deadline, opposite the bank's deadline with no threshold.
  • Neither limb leaves a public trace. The Board of Governors, which collects the notifications, states in its own information-collection filing that no data collected are published.

The interagency final rule runs twenty-one pages in the Federal Register: Document No. 2021-25510, 86 FR 66424, published 23 November 2021, amending 12 CFR parts 53, 225 and 304, effective 1 April 2022 with a compliance date of 1 May 2022. The Government Publishing Office's rendition of the same issue carries it beside the Federal Register page, and the operative sections survive in the codified text of Part 53 — §53.2 for the definitions, §53.3 for the bank, §53.4 for the vendor.

Neither duty runs from an outage. Both run from a judgement.

The rule builds its trigger in two storeys. A "computer-security incident" is an occurrence "that results in actual harm to the confidentiality, integrity, or availability of an information system or the information that the system processes, stores, or transmits" — the 2025 annual edition of 12 CFR Part 53 prints that at §53.2, and the Cornell Legal Information Institute's reproduction of the same section carries it word for word, which corroborates the wording across two reproductions without settling anything about authenticity. An incident becomes a "notification incident" when it "has materially disrupted or degraded, or is reasonably likely to materially disrupt or degrade" one of three things: the bank's ability to carry out banking operations, activities or processes, or to deliver banking products and services to a material portion of its customer base in the ordinary course of business; a business line whose failure would cause a material loss of revenue, profit or franchise value; or operations whose failure or discontinuance would pose a threat to the financial stability of the United States.

Two of those clauses belong to the bank's own commercial judgement. The third belongs to somebody else's estimate of systemic risk. And the definition carries two verbs — has disrupted, or is reasonably likely to — so one limb is a fact and the other is a forecast.

Then the clock. Section 53.3 directs the banking organization to notify "the appropriate OCC supervisory office, or OCC-designated point of contact" by email, telephone or another method the OCC prescribes, and then fixes the interval: the OCC "must receive this notification ... as soon as possible and no later than 36 hours after the banking organization determines that a notification incident has occurred." The FDIC's parallel, 12 CFR 304.23, reads the same way with its own addressee, and OCC Bulletin 2021-55 compresses it into one sentence — 36 hours after the bank determines.

Read the grammar twice. The 36 hours measures nothing about the incident: not the outage, not the damage, not the number of customers behind it. It measures the distance between two acts of the bank — the moment it decides the event qualifies, and the moment its regulator receives the message. Everything before that decision sits outside the interval.

The agencies said as much in the preamble, and said it without a number. They "do not expect that a banking organization would typically be able to determine that a notification incident has occurred immediately upon becoming aware of a computer-security incident", and they "anticipate that a banking organization would take a reasonable amount of time to determine that it has experienced a notification incident".

Then the sentence that matters: "Only once the banking organization has made such a determination would the 36-hour timeframe begin." The rule calls 36 hours reasonable because the notification "does not require an assessment or analysis" — the clock is calibrated to the size of the message, not to the size of the event. And "reasonable amount of time" is given no hours anywhere in the text read. A bank that spends three days reaching its determination has not breached a printed deadline, because no printed deadline is yet running.

The direction of the drafting is worth keeping. Against the January 2021 proposal, the final rule swapped "believes in good faith" for "determines", swapped "could" for "reasonably likely to", deleted a prong that would have captured incidents violating internal policies, narrowed the incident definition from "actual or potential harm" to "actual harm", and excluded designated financial market utilities. Every one of those moves narrows what a bank has to report. Not one of them touches who decides.

Now the vendor. Section 53.4 requires a bank service provider to notify "at least one bank-designated point of contact at each affected banking organization customer" as soon as possible when the provider determines that it has experienced a computer-security incident that "has materially disrupted or degraded, or is reasonably likely to materially disrupt or degrade, covered services provided to such banking organization for four or more hours." Where no such contact was previously supplied, the notice runs to the customer's chief executive officer and chief information officer, "or two individuals of comparable responsibilities", by any reasonable means. Scheduled maintenance, testing and software updates already communicated to the customer are carved out.

The two limbs are cut from different material, and the mirror holds. The bank has a deadline with no threshold: an hour of disruption, if the bank judges it material, is enough to start 36 hours running. The vendor has a threshold with no deadline: three hours and fifty minutes of degraded covered service produces no duty at all under this text, and four hours produces a duty whose only temporal phrase is "as soon as possible".

That is not an indictment of the vendor clause. It is the better-drafted of the two, precisely because of the threshold: it spares a provider from reporting every blip and spares its customers a flood of noise, and it names human beings — a designated contact, or the chief executive and the chief information officer — rather than a supervisory office. The asymmetry lies in what each limb can measure from outside. A regulator can reconstruct a 36-hour clock from two dates.

A four-hour test needs a start, an end and a judgement that the degradation was material, and the rule assigns none of those three to anyone outside the provider. "As soon as possible" has no denominator in the text; it is measured against the provider's own capacity, which is a different kind of obligation from a ceiling.

Then the question that decides whether any of it can be checked. The Board of Governors, which collects the notification data under FR 2231, filed a supporting statement for that information collection with the Office of Management and Budget — the current three-year extension, dated 13 May 2025 — and one line in it is flat: "No data collected by this information collection are published." The same statement describes what is filed as "generally nonpublic commercial or financial information, which is both customarily and actually treated as private by the respondent", kept confidential under exemption 4 of the Freedom of Information Act (5 U.S.C. 552(b)(4)), with exemption 8 noted for material contained in or related to examination reports, and it records that service-provider notices to banking organization customers generally are not provided to the Board at all. "Generally", twice, is doing real work: this is a described practice of withholding, not a statutory seal, and the filing does not claim otherwise.

Bound that finding to the material. Across the eight documents read for this piece on 19 November 2026 — the final rule, its Government Publishing Office rendition, the 2025 Part 53 text, the Cornell reproduction of §53.2, the FDIC's Part 304 codification, the Board's FR 2231 statement, OCC Bulletin 2021-55 and the January 2021 proposal — no provision or statement requiring publication of an individual notification was located, and no per-notification public identifier was located. Nor was a first-party statement publishing an aggregate count.

The only figure of that shape in the pack is the impact analysis's estimate of approximately 150 notification incidents a year, and that is an estimate of what the agencies expected, not a count of what arrived.

The delta, measured the same day. A literal search for "Federal Reserve" returned two published rows on this desk; the nearer is a Case File from 3 November 2026 on the Board's own emergency exercises and the incomplete per-system recovery plans its supervision found — internal continuity, the institution rehearsing its own failure. A literal search for "incident notification" returned nothing indexed. This piece reads the opposite direction of the same relationship: the duty a supervisor imposes on the institutions it examines, and the two unlike instruments it uses to time them.

Boundary. The 36-hour duty quoted above comes from 12 CFR 53.3, from the FDIC's parallel at 12 CFR 304.23 and from the interagency rule's own text, including the preamble passage on determination that appears in the Federal Register's published page. The Board's own codified paragraphs were not read for this piece. The confidentiality finding rests on the Board's filing and its two uses of "generally", not on a reading of the final rule's confidentiality discussion, and it says nothing about the OCC's or the FDIC's handling.

Whether a particular bank has ever reached its determination late, or a particular provider has ever measured four hours generously, is not recorded in anything read here.