The CVE Numbering Authority (CNA) Operational Rules are the CVE Program's own governing document for how a CVE ID and its CVE Record actually get assigned, published and corrected — not just the familiar "CVE-YYYY-NNNNN" format, but the whole process behind it. The current version, 4.2.1, was approved on 14 October 2026 and took effect on 20 October 2026, after version 4.2.0 (effective 25 August 2026) replaced 4.1.0 (effective 14 May 2025). Per the rules' own Introduction, a CNA "is an organization authorized to assign CVE IDs and publish CVE Records, ideally as part of the initial Public Disclosure of a Vulnerability" — and the obligations that come with that authorization are considerably more specific than most CVE consumers realize.
Key takeaways
- The program is a hierarchy: Top-Level Roots and Roots manage CNAs, and CNAs of Last Resort catch what no scoped CNA covers.
- Two releases in 2026 changed the governance body, the problem-type rule and the vulnerability-determination numbering.
- Each CNA's Root or TL-Root approves its Scope Definition, and the CNA may assign only inside it.
- The CNA with the best scope gets 72 hours to accept or decline before a Root steps in.
How is the CVE Program organised: Root, TL-Root, CNA, CNA-LR?
The program is a hierarchy: Top-Level Roots and Roots manage CNAs, and CNAs of Last Resort catch what no scoped CNA covers. Roots oversee a set of CNAs and, under 2.2.1, "MAY make additions to the CNA Rules" for their own hierarchy, with TL-Root approval; the additions "MUST NOT conflict with the primary CNA Operational Rules." Top-Level Roots (TL-Roots) sit above Roots. Ordinary CNAs assign CVE IDs within their own agreed Scope Definition. CNAs of Last Resort (CNA-LRs) handle overflow and first-refusal cases; the rules note they "typically have broader scopes than CNAs" and "MAY limit effort to optimize service given resource constraints, for example, by not notifying Suppliers who are not CNAs." MITRE holds several roles at once: per 3.1.12.1, "MITRE currently acts as the Secretariat, a TL-Root, and a CNA-LR."
What changed between CNA Rules 4.1.0 and 4.2.1?
Two releases in 2026 changed the governance body, the problem-type rule and the vulnerability-determination numbering. Comparing the 4.2.1 rules with 4.1.0:
| Topic | 4.1.0 (May 2025) | 4.2.1 (October 2026) |
|---|---|---|
| Who manages the rules and decides scope disputes | CVE Board | CVE Program Executive Committee (ESG); the change came with 4.2.1, as 4.2.0 still names the Board |
| Problem type in a record (5.1.7) | "MUST identify the type of Vulnerability" | "MUST specify the type or cause", with 5.1.7.1 allowing "not enough information" only until the CNA can determine it |
| Insecure defaults (4.1.3 in 4.2.1) | One sentence | Expanded: CNAs "MUST make reasonable efforts to understand default configuration settings" |
| Clause numbers in 4.1 | 4.1.1–4.1.15 | 4.1.1–4.1.14 (malicious code is now 4.1.7, dependency updates 4.1.11, EOL 4.1.12) |
Publishing deadlines and the first-refusal model did not change. Rule changes "MUST NOT be applied retroactively" to existing records (2.1.5).
Scope Definitions: who decides who can assign what?
Each CNA's Root or TL-Root approves its Scope Definition, and the CNA may assign only inside it. "CNAs MUST develop, document, and operate within a Scope Definition that describes the types of Vulnerabilities for which the CNA will assign CVE IDs" (3.1.1); the scope is published in the List of Partners. Under 3.1.3 a Supplier CNA "SHOULD define Scope in reasonably broad terms": all of its products, a named product list, or a covered-and-excluded list. 3.1.11 lets a CNA state whether it assigns IDs for End-of-Life products, which matters for device makers with long-lived installed bases. Scope disputes go through the Root, TL-Root or Council of Roots first; if unresolved, the Secretariat brings them to the ESG, which "makes the final decision" (3.1.10).
How does the 72-hour first-refusal clock work?
The CNA with the best scope gets 72 hours to accept or decline before a Root steps in. CVE ID assignment follows "a 'right of first refusal' model" (4.2.1). Under 4.2.1.2, for a vulnerability that is already Publicly Disclosed, three triggers hand assignment elsewhere: the scoped CNA preemptively declines, declines within 72 hours, or doesn't respond within 72 hours. Then "an appropriate Root MUST make a Vulnerability determination" and direct another CNA or a CNA-LR to assign "no later than 72 hours after becoming aware of the first refusal." For a vulnerability that is still undisclosed, the requester follows the dispute guidance in 4.6 instead (4.2.1.3). A CNA whose intake form imposes restrictive terms "effectively waives their right of first refusal" (3.2.4.2).
What are the publishing deadlines?
The record should follow public disclosure within 24 hours and must follow within 72.
| Rule | Requirement level | Deadline |
|---|---|---|
| 4.5.1.3 | SHOULD | Publish the CVE Record within 24 hours of Publicly Disclosing an assigned CVE ID |
| 4.5.1.4 | MUST | Publish within 72 hours of Publicly Disclosing an assigned CVE ID; after that a Root may direct a CNA-LR to publish |
| 4.5.1.6 | SHOULD | Publish within 72 hours of learning a third party already disclosed an assigned CVE ID |
| 4.5.3.5 | MUST | Reject any unused or unpublished CVE ID |
What must a CVE Record contain in CVE JSON 5.2?
Section 5.1 sets the minimum, and the CVE Record Format 5.2.0 schema enforces most of it. In 5.2.0 a published record's CNA container requires providerMetadata, descriptions, affected and references; each affected item needs vendor and product (or a package collection and name) plus versions or a defaultStatus. Version 5.2.0 also added packageURL for purl identifiers.
| Rule (4.2.1) | Requirement | CVE JSON field |
|---|---|---|
| 5.1.3, 5.1.4 | At least one affected Product, marked "affected" or "unknown" | affected[] with vendor, product, versions[].status |
| 5.1.7 | Type or cause; CWE recommended | problemTypes[].descriptions[].cweId |
| 5.1.9, 5.2.5 | Prose description, at least one in English | descriptions[] with lang: "en" |
| 5.1.10, 5.3.1 | At least one public reference that exists before or with the record | references[].url |
| 5.2.4 | No credit in the description | credits[] instead |
The record below is synthetic: the ID, organisation UUID, vendor and URL are invented example values. It validates against the 5.2.0 schema.
{
"dataType": "CVE_RECORD", "dataVersion": "5.2",
"cveMetadata": { "cveId": "CVE-2026-0000", "assignerOrgId": "00000000-0000-4000-8000-000000000000", "state": "PUBLISHED" },
"containers": { "cna": {
"providerMetadata": { "orgId": "00000000-0000-4000-8000-000000000000" },
"descriptions": [ { "lang": "en", "value": "Out-of-bounds write in the DHCP client of ExampleCo X200 gateway firmware before 2.4.1 allows a device on the same LAN segment to execute code via a crafted DHCP offer." } ],
"affected": [ { "vendor": "ExampleCo", "product": "X200 gateway firmware", "defaultStatus": "unaffected",
"versions": [ { "version": "2.0.0", "lessThan": "2.4.1", "status": "affected", "versionType": "semver" } ] } ],
"problemTypes": [ { "descriptions": [ { "lang": "en", "type": "CWE", "cweId": "CWE-787", "description": "CWE-787 Out-of-bounds Write" } ] } ],
"references": [ { "url": "https://example.com/security/ESA-2026-001", "tags": ["vendor-advisory"] } ]
} }
}
The description names product, fixed version, attacker position and impact, and leaves out the researcher's name. CWE choice is covered in the CWE Top 25 guide; a CVSS vector for the optional metrics block can be built with the CVSS calculator.
When does a firmware issue count as a vulnerability?
The rules leave room for judgment but draw hard lines for some cases device makers meet often. "CNAs are expected to exercise discretion and judgment, especially when making complicated Vulnerability determination (4.1) and CVE ID assignment (4.2) decisions" (4.4). When a CNA cannot reach a clear call, it "SHOULD err on the side of assignment and SHOULD assign CVE IDs," which "allows CVE users to identify and discuss the 'Vulnerability-like' issue."
For firmware teams, four clauses matter most:
- Insecure defaults count. "CNAs SHOULD determine insecure default configuration settings to be vulnerabilities" (4.1.3); a shared default password or an open debug service is in scope.
- Updating a component is not a new vulnerability. "Updating a library to address a Vulnerability in that library MUST NOT be determined to be a new Vulnerability in a Product that uses the library" (4.1.11); the product advisory "SHOULD reference the CVE ID for the Vulnerability in the library."
- End-of-Life alone is not a vulnerability (4.1.12), and malicious code in general is not either (4.1.7).
- One vulnerability, one ID. A bug shared by supported and EOL versions gets one CVE ID (4.2.17.8).
How does a device maker become a CNA?
It applies through the CVE Program's web form, meets three requirements and completes onboarding; per the cve.org Partner page, "There is no monetary fee" and "There is no contract to sign." The requirements are a public vulnerability disclosure policy, a public source for new vulnerability disclosures, and agreement to the CVE Terms of Use. The page lists five steps:
Contactthe CVE Program. Registerwith the "Register to Become a CNA" form: organisation details, estimated annual number of CVE IDs, expected scope, disclosure policy and advisory links. Meet requirements(the three above). Onboard: watch the CVE Lightning Talk video, attend an introductory session and complete practice examples. Join: the organisation is added to the List of Partners and announced.
The partner data behind cve.org listed 567 organisations on 11 November 2026, 460 of them typed "Vendor" (CNAsList.json). Records go through the submission process the program designates (4.5.1.2); the 5.2.0 release notes name Vulnogram as a widely used open-source tool for creating and submitting CVE Records.
Should a device maker become a CNA at all?
Become one if you can staff the clocks; otherwise let a Root or CNA-LR assign for you.
Check staffing firstThe 72-hour first-refusal and publication clocks are operational commitments; see PSIRT response SLA examples for realistic targets. Scope broadly3.1.3 favours "all of a Supplier's Products" over a list that needs updating with every launch. Automate the publication clockThe gap between "SHOULD within 24 hours" and "MUST within 72 hours" is where most PSIRTs under-provision. Keep a public distribution pointmeeting 4.5.2.2 and 4.5.2.4: no login, no restrictive terms. It doubles as the record's public reference under 5.3. Pair CNA status with a disclosure policy(template); 3.2.6.1 requires CNAs to publish how they assign IDs within their disclosure process. For setting up the team itself, see PSIRT setup for a hardware company.
How do the CVE licence terms affect reuse?
CVE Records are published under the CVE Program's own Terms of Use, catalogued by SPDX as cve-tou. The grant is a perpetual, worldwide, royalty-free, irrevocable copyright licence to reproduce, create derivative works of, publicly display and distribute the CVE List, conditioned on reproducing MITRE's copyright designation and the licence terms.
Who governs the CVE Program today?
MITRE runs the program as Secretariat under CISA sponsorship, and the 4.2.1 rules give rule-making and final scope decisions to the CVE Program Executive Committee (ESG). In April 2025 a lapsing CISA contract briefly put the program's continuity in question before CISA extended MITRE's contract by 11 months; CVE Board members also launched the separate CVE Foundation at that time. Records already published are unaffected by rule changes (2.1.5).
How does this fit with vulnerability disclosure more broadly?
The CNA rules govern identifier assignment, not disclosure policy: "The CVE Program itself does not follow or require a specific Vulnerability disclosure policy" (3.2.6). A PSIRT still needs its own coordinated disclosure process, which is where ISO/IEC 29147 and 30111 and the FIRST PSIRT Services Framework apply, alongside a vulnerability disclosure policy and, for CRA-scoped products sold in the EU, the 24-hour/72-hour CRA reporting obligations, which run on a separate clock. Machine-readable advisories that reference CVE IDs are covered in CSAF/VEX for device makers, and NVD enrichment delays in the NVD backlog guide. Becoming a CNA is a separate, voluntary step: a supplier can run a compliant PSIRT without it.
Which part can a vulnerability-matching tool help with?
Matching tools help with the inbound side, not with assigning or publishing CVE IDs. SoftWSP extracts a CycloneDX 1.6 SBOM from firmware, matches its components against NVD, CISA KEV, FIRST EPSS and CISA CSAF ICS advisories, and has AI draft a cited report that an engineer reviews; watchlist alerts fire on a CVE ID or a vendor and product. That tells a PSIRT which third-party CVEs reach its products, the case 4.1.11 covers. SoftWSP does not assign CVE IDs, write CVE JSON or VEX, or submit records. See PSIRT services.
This guide quotes the CVE Program's CNA Operational Rules 4.2.1, the cve.org Partner page, the CVE Record Format 5.2.0 schema and the SPDX catalogue entry for the CVE Terms of Use. It is not legal advice.
Put this guide to work
Frequently asked questions
What exactly is a CVE Numbering Authority (CNA)?
Per the CVE Program's own CNA Operational Rules: "a CVE Numbering Authority (CNA) is an organization authorized to assign CVE IDs and publish CVE Records, ideally as part of the initial Public Disclosure of a Vulnerability." The rules name typical CNA types explicitly: "Software and hardware Suppliers, including product security incident response teams (PSIRTs); Open source software projects, maintainers, and foundations; Service providers; Vulnerability researchers; National computer security incident response teams (CSIRTs); Vulnerability coordinators; Vulnerability databases; Bug bounty providers." Most Supplier CNAs assign CVE IDs for vulnerabilities in products they own, develop or maintain — but a CNA's rights are limited to its own agreed-upon **Scope Definition**, not all vulnerabilities everywhere.
If my company reports a vulnerability but the vendor won't assign a CVE ID, what happens?
The rules describe a "right of first refusal" model (4.2.1): "The CNA with the most appropriate scope MUST be contacted and given the first opportunity to assign. In many cases, this CNA is the Supplier of the Product in question." For an already Publicly Disclosed vulnerability, if that CNA "preemptively documents that it will not assign, or responds within 72 hours that it will not assign, or does not respond within 72 hours, then an appropriate Root MUST make a Vulnerability determination" and direct another CNA or CNA-LR to assign "as quickly as possible and no later than 72 hours after becoming aware of the first refusal." For vulnerabilities that are not Publicly Disclosed, the rules instead require following the Program's formal CVE Record dispute process rather than defaulting to another assigner.
How fast does a CNA have to publish the actual CVE Record, once a CVE ID is Publicly Disclosed?
The rules set two deadlines, one aspirational and one mandatory: CNAs "SHOULD publish a CVE Record to the CVE List within 24 hours of Publicly Disclosing a CVE ID assigned by the CNA" (4.5.1.3), and separately "MUST publish a CVE Record to the CVE List within 72 hours of Publicly Disclosing a CVE ID" (4.5.1.4). Missing the 72-hour mandatory deadline lets the CNA's Root direct an appropriate CNA-LR to publish the record instead, with ownership of the record potentially transferring away from the original CNA.
What must a CVE Record actually contain, at minimum?
Per Clause 5.1, a CVE Record "MUST identify at least one affected Product using information such as Supplier and Product names, versions, and dates," "MUST specify the type or cause of the Vulnerability" (the record "SHOULD use the Common Weakness Enumeration (CWE) to classify the type or cause"), "MUST contain a prose description," and "MUST contain at least one public reference" that "MUST NOT be a reference to the CVE Record itself" — meaning a CVE Record can never be the first public disclosure of a vulnerability; some other public reference has to exist first or at the same time (5.3.1). The description itself "MUST NOT credit people, organizations, or tools" (5.2.4) — researcher credit goes in the CNA's own advisory, not the CVE Record's description field.
Sources
- CVE Program: CVE Numbering Authority (CNA) Operational Rules, Version 4.1.0 (effective and approved by the CVE Board 14 May 2025; superseded by 4.2.0 and 4.2.1)
- CVE Program: "CNA Operational Rules Version 4.1.0" Now in Effect (official CVE Program announcement of the v4.0 -> v4.1.0 transition, no breaking changes)
- SPDX License List: Common Vulnerability Enumeration ToU License (cve-tou) — catalogued summary of the CVE Program's Terms of Use licence terms
- The Record (Recorded Future News): CISA extends CVE program contract with MITRE for 11 months amid alarm over potential lapse (April 2025 funding crisis that prompted the CVE Board to form the CVE Foundation)
- CVE Program: CNA Operational Rules, Version 4.2.1 (effective 20 October 2026) — the rules text quoted in this guide
- CVE Program: CNA Operational Rules, Version 4.2.0 (approved 20 August 2026, effective 25 August 2026)
- CVE Program: Partner page (CNA requirements, no fee or contract, five-step onboarding process)
- CVE Program website data: CNAsList.json (partner list behind cve.org, CVEProject/cve-website)
- CVEProject/cve-schema: CVE Record Format version 5.2.0 release (29 October 2025; packageURL added)
- CVEProject/cve-schema: CVE Record Format 5.2.0 JSON schema
- Vulnogram (open-source CVE JSON editor and submission tool)
About this guide
SoftWSP Firmware Security Desk
Covers the PSIRT & disclosure cluster: 15 guides.

In this cluster
Questions about this page?
