Summary

  • CISA's Known Exploited Vulnerabilities Catalog prints one Due Date per entry and attributes it to BOD 26-04, which replaced a flat two-class rule with a four-variable, sixteen-row remediation matrix.
  • The directive starts the clock at whichever comes first — the KEV listing, or the agency's own identification of the vulnerability on an asset plus its CDM dashboard update — and gives the final asset-exposure determination to the agency. The catalog computes its date from CISA's inputs and its own anchor, and publishes nothing about the other clock.

On 18 November 2026 the Known Exploited Vulnerabilities Catalog did what it does most weekdays. It took in vulnerabilities that attackers are already using and printed a deadline beside each one. CVE-2026-90436 arrived with a Due Date of 2026-11-21 and a line reading "Forensic triage required per BOD 26-04: Yes." Three calendar days. CVE-2026-40217 arrived the same day with a Due Date of 2026-12-02 and triage "No." Fourteen calendar days. Same catalog, same day of admission, fourteen days against three, and the row does not say why.

The reason is not in the row. It is in a document the row cites.

The pointer goes somewhere narrow

"Per BOD 26-04" is not decoration. It points to BOD 26-04: Prioritizing Security Updates Based on Risk, dated 10 June 2026, which supersedes and revokes BOD 19-02 and BOD 22-01 — the directive whose own page now carries the word revoked in its title.

BOD 22-01 was readable off the row: six months for a vulnerability whose CVE ID was assigned before 2021, two weeks for everything else, adjustable in cases of grave risk to the Federal Enterprise. Two classes, one of them guessable from the CVE number alone. The successor replaced that with a computation.

BOD 26-04 states that remediation urgency is determined by four variables — Asset Exposure, KEV Status, Exploit Automation and Technical Impact — and sends agencies to Table 1, the Vulnerability Response Timeline. Asset Exposure asks whether the vulnerable asset is publicly exposed. KEV Status asks whether the CVE sits in the catalog. Exploit Automation asks whether an adversary can automate all the steps needed to exploit it. Technical Impact is partial or total control. Days in the table are calendar days. A local note is needed here before the arithmetic, because it governs everything that follows: on the directive page Table 1 is published as an image, its caption and variable definitions readable, its cells not. The row mapping below comes from a reproduction published by Datadog, which I read and name as a secondary reproduction because CISA's own table is not text on the page and the directive states no row values in prose.

That reproduction gives Table 1 sixteen rows and five outcomes: three days plus forensic triage, three days, fourteen days, sixty days, and "fix on system upgrade," which the directive defines as remediation at the next scheduled major upgrade or rebuild unless conditions change.

The arithmetic that never gets printed

Sixteen rows look like a rich instrument. They collapse against one input the catalog fixes for you.

Every row in the catalog has KEV Status = yes. That is what being in the catalog means. Sweep the eight rows with KEV Status = yes and only three outcomes survive: three days with forensic triage, three days, or fourteen days. The sixty-day outcome and the fix-on-upgrade outcome are unreachable for any vulnerability CISA has catalogued. The catalogue's entire published range is three to fourteen calendar days — which is exactly what the two entries above show, and no part of the row tells you the range has a ceiling.

Sweep further and the triage flag becomes more interesting. Exactly three of the sixteen rows carry it. All three share Technical Impact = total and are distinguished by the second condition that was also true: publicly exposed, or automatable by the adversary. So the one input the catalog does publish — that Yes/No line — is a two-condition conjunction compressed into a word. It tells a reader that something beyond total impact held, and withholds which of the two things it was. A defender reading "triage: Yes" learns that the deadline is three days.

She does not learn whether the reason was her exposure or the adversary's automation, and those two facts call for different work: one is a perimeter question, the other a patch-cadence question.

Two clocks, one printed

Here is the layer that matters most, and it is in the directive's own timing rule.

BOD 26-04 states that the Table 1 timelines begin when either CISA adds the vulnerability to the KEV Catalog or, pursuant to BOD 23-01, the agency enumerates or identifies the vulnerability on an asset and updates the CDM Program Agency and Federal Dashboard — whichever event occurs first.

Read that twice. For a catalogued vulnerability there are two possible starts. One is the listing, which is public and which the row prints as Date Added. The other is the agency's own identification of the same vulnerability on its own asset, followed by a dashboard update, which is not public and which the row does not print. The directive picks the earlier of the two. The catalog computes its date from the one it can see.

So the printed Due Date can be a later date than the deadline the agency's own clock implies. If a federal agency found the vulnerability on its estate before CISA catalogued it, its three or fourteen or sixty days began at the earlier event, and the number it must beat is not the number printed beside the CVE. The catalog is not wrong to publish a date. It is publishing a date computed from a start it can observe, against a rule whose start it cannot.

The input that is computed twice

The second divergence is in the variable, not the anchor. CISA's implementation guidance for BOD 26-04 states that agencies should remediate by the "dueDate" as determined by the Vulnerability Response Timeline table, and that in determining the catalog due date CISA calculates the date according to that table using available data to determine whether there is evidence that the vulnerability is present in a publicly exposed asset. It then gives the worked case: exposed, total technical impact and automatable yields a three-day deadline.

The same guidance gives the final asset-exposure determination to the agency, and states that remediation timelines are determined asset by asset. It names where the inputs come from: CISA's Vulnrichment Program supplies KEV status, automation and technical impact; exposure evidence may come from CISA Cyber Hygiene or from third-party scanners.

There are therefore two exposure findings in play — CISA's, taken from evidence it holds about the federal estate in aggregate, and the agency's, taken about a specific box. Only the first shapes the printed date. The second shapes the obligation. And the directive makes the consequence explicit: timelines are dynamic, because removing a system from the internet can move the required timeline back while adding a vulnerability to KEV can move it forward.

That is not a scandal. Asset-by-asset determination is the only defensible way to assign an asset-by-asset deadline, and a central catalog cannot see an agency's firewall rules. But it does mean the published number is a computed estimate of a deadline, published by the party that does not own the asset, from a variable the party that owns the asset has final authority over. The catalog shows the arithmetic's output and not its inputs.

Nothing about this is softened by the fallback rules. The guidance states that CISA does not issue waivers or exceptions for actions required in cyber directives, that agencies with special use cases should contact CISA's directives office, and that an asset which cannot be remediated within the timeframe must be removed from the agency network. There is no published exception to inspect because there are no exceptions to publish — the relief valve is disconnection, and disconnection is not a record either.

What the row cannot tell you

Now the part a stranger would look for. Did the thing get fixed?

The catalog does not say. It prints the CVE, the name and description, related CWE identifiers, a "Known To Be Used in Ransomware Campaigns?" line worth Known or Unknown, the triage flag, Date Added, Due Date and, for some entries, a required action. It does not publish per-entry remediation status, agency completion, waivers, extensions or an overdue indicator, and I found none in the blocks I read.

The status goes elsewhere. Phase I Required Action 3 requires agencies to automate reporting on the status of KEV-listed vulnerabilities through the CDM Dashboard, and requires agencies not fully automated to report manually every two weeks. Phase III Required Action 7(a) separately requires agencies not fully automated through CDM to report asset-tagging information to CISA every seven days in a machine-readable format.

CISA Action 9 requires CISA, by the end of each fiscal year, to give a status report to the Secretary of Homeland Security, the Director of OMB and the National Cyber Director identifying cross-agency status and outstanding implementation issues. None of those limbs asks anyone to publish a per-vulnerability outcome, and the directive does not require publication of per-agency compliance results.

CISA's own catalog repository closes the loop in the least convenient place: it acknowledges that the KEV file has no inline revision history or log that is easily accessible after the fact. So a row that is later corrected — a due date adjusted after new exposure evidence, a triage flag flipped — does not leave a public trace at the row. The current state is the only state.

That combination has a specific consequence worth stating plainly. The catalog's due date is a live obligation with a hard number attached, and the record that would let an outsider test whether the number was met is not published at the row, is not published in aggregate per agency, and is not reconstructable later because the row has no history. The number is auditable. Its performance is not.

There is a counterweight, and it is real. This catalog is more instrumented than most registers I read. It publishes a date that binds, a flag that means something, and a ransomware line; it carries inclusion criteria — an assigned CVE ID, reliable evidence of active exploitation in the wild, and a clear remediation action, as CISA's catalog guidance page restates them; and it removed its own escape hatch by refusing waivers. BOD 22-01's rule was legible off the row and required almost no literacy to check. BOD 26-04's rule is better at matching urgency to risk and worse at being checked, and those are two separate properties that a single table cannot hold at once.

What would make the row checkable

Three fields, and one page.

Print the inputs beside the output: the exposure finding CISA used, the automation and technical-impact values the calculation ran on, and the date the calculation was made. The row already shows the answer; showing the operands costs nothing that the catalog is currently protecting and turns a number into an argument a reader can follow.

Print the agency's own determination as a distinct record, not as a replacement. The guidance already establishes that the asset-specific finding governs; publishing it beside the catalog-level one would show a reader where the two diverge instead of leaving that divergence as my inference from two documents.

And publish a revision log for the row. CISA's repository states there is none accessible after the fact. A date that can change and leave no trace is not a record; it is a status display.

The delta

I have written about this agency twice before. My earlier CIRCIA briefing counted two printed incident clocks against a manner-and-form rule still at OIRA; an earlier piece read the Cyber Safety Review Board. Neither read a register row against the directive that supplies its number, and neither separated a catalog-level exposure inference from an agency's asset-specific determination. That is the gap this piece fills, and it is a gap in the instrument rather than in the agency's conduct: nothing here asserts that any agency missed any deadline, that any determination was wrong, or that CISA miscomputed any date it published.

The Cybersecurity and Infrastructure Security Agency built a catalog that tells you exactly how long you have and declines to tell you how it decided. Sixteen rows became one date. The date is right, in the sense that it follows from the table. It is just not, necessarily, yours.