Summary

  • The published record format carries the record state at cveMetadata.state, and its two specialised definitions each hard-code one allowed value: PUBLISHED or REJECTED. No disputed state exists in it.
  • A contest is carried as a container tag plus a free-text reason. The dispute policy names the adjudicator and prints three business-day clocks, and requires that tag and reason to be removed once the dispute is resolved or deemed invalid.

The state field is the smallest instrument in vulnerability management and the one everything downstream leans on. A scanner, a patch queue, a regulator's filing, a board's risk register: each of them reads a state before it reads a sentence. So it is worth opening the field itself rather than the prose around it.

In the published record format the field sits at cveMetadata.state, and it is not a vocabulary — it is two closed lists. The published metadata definition prints "enum": ["PUBLISHED"]; the rejected one prints "enum": ["REJECTED"]. The description above them is five words: "State of CVE - PUBLISHED, REJECTED." (CVE Record Format schema) The same schema tells you what a revision is: a counter that "start[s] at 1, and increment[s] by 1 each time a submission from a data provider changes this CVE Record", moving between the rejected and published shapes on a PUBLISHED->REJECTED or REJECTED->PUBLISHED transition. Who made the change, and when, are not in that counter.

A contest lives somewhere else entirely. The generated documentation for the format prints the CNA tag registry as three values — unsupported-when-assigned, exclusively-hosted-service, disputed — and defines the last one this way: "When one party disagrees with another party's assertion that a particular issue in software is a vulnerability, a CVE Record assigned to that issue may be tagged as being 'disputed'." (generated schema documentation) That is a tag inside a container's tag array, not a state. The entry read prints no adjudicator for it, no decision date, no basis of decision, no resolution field.

What governs a contest is a separate document, and it is more forthcoming than the record. The CVE Record Dispute Policy prints "Version: 2.0.0", "Approved date: July 2, 2025", "Effective date: July 2, 2025", and states that it is "enforced by Top-Level Roots (TL-Root) and Roots" and applies to records disputed after that date. It names the deciders plainly: "CNAs, Roots, and TL-Roots MAY serve as CVE Record dispute Adjudicators when necessary." Disputes run "through the appropriate (TL-) Root hierarchy, beginning with the CNA responsible for the affected scope"; escalation goes to a Root or TL-Root, whose decisions are "final, except in cases involving cross-hierarchy scope issues"; disputes spanning hierarchies "will be adjudicated by the Council of Roots."

It also prints clocks, and they are real ones. The adjudicator "MUST acknowledge receipt and initiation of the dispute in writing within three business days", must "reach a decision within five business days after the three-day acknowledgment period", and "If an extension exceeds 15 business days, any involved party MAY escalate the dispute to the Root". Three plus five gives eight business days to the outside of the printed decision window; that is my arithmetic on the policy's own words, not a figure the policy prints.

I do not convert the fifteen-day extension into a date, because the policy prints no start point for it and no holiday calendar.

Then comes the sentence that decides this article. "If the dispute is later deemed invalid or resolved, the Adjudicator MUST remove the disputed tag and reason." Set that beside the purpose the same policy declares for the tag: "Placing the disputed tag allows consumers to determine whether there has been a dispute for a record." The instrument whose stated job is to tell consumers a dispute existed is ordered deleted at the moment the dispute stops. Not amended, not stamped closed — removed, with nothing printed about who removed it, on what date, or on what finding.

Ongoing disputes stay tagged; "There are cases where there will be ongoing disputes after a review. In that case, the CVE Record will continue to be tagged as disputed and will continue to provide a reason for the dispute."

The Program's own explainer goes further than the record can carry. Published under the Council of Roots' byline on 16 December 2025, CVE Record Disputes Explained says: "When a dispute arises, the CVE Record is updated to include both positions. Each side's rationale is documented, and the record itself is explicitly tagged as 'disputed.'" And, of the end state: "Unless there is a convergence of perspective, the record may remain in this state indefinitely." Documenting each side's rationale is a stronger promise than the format keeps: on the entry read, the record holds one tag and one reason.

The rules layer routes rather than decides. CNA Operational Rules v4.2.6 — "Version: 4.2.6", approved 14 October 2026, effective 20 October 2026 — makes the dispute policy mandatory for "[d]isagreements with a CNA's decision or judgment" and, in section 4.6, permits a dispute, asks CNAs to try to resolve it first, and incorporates the policy by reference. Section 4.6 itself names no adjudicator, no evidence standard, no time limit. Rule 4.5.3.10 requires the tag "and ... an explanation in the CVE Record"; rule 4.5.3.11 requires Program members to present that explanation when rendering record data. The same rules as a web page carry the identical numbering and print no version or effective date.

The older rules show how much moved. CNA Rules v3.0 section C.5 put the contest in the prose — "** DISPUTED **" at the start of the description and a short note at the end — and made the disputing party's link and quote optional: "However, neither are required." That version at least left something human-readable behind. Its successor moved to a tag and then made removing the tag the marker of an ending.

One more disagreement is worth publishing because a stranger will meet both pages. The CVE Program Glossary says "A CVE Record can be in one of the following states: Reserved... Published... Rejected...", glosses Reserved as "the initial state of a CVE ID", and keeps Rejected on the CVE List as an identifier users should no longer use. The schema exposes two values. Three named states against two enumerations: a difference of one, counted from the pages themselves. Neither page prints a version or a date, so the divergence cannot be dated from either. Two printed dates can be subtracted, though: the policy took effect on 2 July 2025 and the explainer is dated 16 December 2025, 167 days; the current operational rules took effect on 20 October 2026, 475 days after the policy.

This desk has read the identifier's meaning before — a number that coordinates disclosure and never proves a repair, one numbering authority's workload, the 1999 launch and its 321 entries. None of that work reads the state field as an instrument, asks who clears a contest, or sets the policy against the schema. That is the delta here: not what a CVE ID means, but what a registry can prove about the argument it recorded.

The MITRE Corporation, which operates the CVE Program, is listed in our directory.

Everything above rests on eight documents read on 21 November 2026: the schema, its generated documentation, the rules in both PDF and web form, the superseded CNA Rules v3.0, the dispute policy, the glossary and the Council of Roots explainer. The dispute policy text obtained may be partial, and the schema read is the repository's main branch, which prints no version inside the quoted block. So the honest statement of the finding is bounded in the same breath as it is made: on the documents read, a stranger can learn that a record was contested, and can learn nothing from the record about how the contest ended.

The fix is a field, not a confession. A record that has been disputed and settled could carry the adjudicator, the decision date, the stated basis and the resulting state, with the tag retained and stamped closed rather than deleted. That would cost one schema revision and make every printed clock checkable after the fact. Until then, the registry keeps an excellent account of the argument and destroys the only page that would have shown the verdict.