# Vulnerability Records > [!abstract] What this note is > The public records a software vulnerability passes through, from first report to working exploit: who keeps each one, what it holds, and what changed in 2025 and 2026. Written at Jeff's request on 13 October 2026 as a companion to [[Exploit Database]], and deliberately conservative: every statement was read from the keeper's own pages on that date, or is marked as coming from a reputable secondary source in the sense the [[Disclaimer]] defines. Nothing is projected. Part of [[AI Security for Users]]. ## The path a vulnerability takes A hole in software becomes a public fact in stages, and a different record holds each stage. 1. **Report.** Open-source disclosures are posted to the oss-security mailing list, hosted by Openwall, whose archive runs from 2008 and held 34,629 messages on 13 October 2026 [1]. Full Disclosure, hosted by the Nmap project at SecLists, describes itself as "a public, vendor-neutral forum for detailed discussion of vulnerabilities and exploitation techniques"; its archive runs from July 2002 [2]. Coordinated programmes such as the Zero Day Initiative, established in 2005, pay researchers for findings, share them privately with the vendor, and publish an advisory after the patch, with a standard window of 120 days [3]. 2. **Identifier.** The CVE Program assigns the CVE number. It launched publicly in September 1999 and is maintained by MITRE under a contract overseen by CISA. Identifiers are assigned by CVE Numbering Authorities, which include vendors such as Microsoft and Red Hat and coordinators such as CERT/CC [4, secondary]. 3. **Description.** NIST's National Vulnerability Database adds a severity score (CVSS), affected-product identifiers (CPE) and a weakness class (CWE) to CVE records [5]. 4. **Priority.** CISA's Known Exploited Vulnerabilities catalogue lists holes "exploited in the wild." It held 1,740 entries on 13 October 2026, each with a date added and a due date, and its entries direct federal agencies to Binding Operational Directive 26-04 [6]. The Exploit Prediction Scoring System, maintained by a special interest group at FIRST with scores generated by Empirical Security, publishes for every CVE a daily estimate of "the probability that a published CVE will be exploited in the wild in the next 30 days" [7]. 5. **Demonstration.** [[Exploit Database]] holds working proofs for each hole. 6. **Behaviour.** MITRE ATT&CK is "a knowledge base of adversarial techniques based on real-world observations": what attackers do, across Enterprise, Mobile and ICS systems, at content version 19.2 in October 2026. It documents observed behaviour, not vulnerabilities [8]. For open-source packages, two records sit beside the CVE path. OSV, an open-source project hosted under Google's GitHub organisation, aggregates advisories across 38 ecosystems in one schema, with 229,597 entries for npm alone on 13 October 2026 [9]. The GitHub Advisory Database publishes advisories in the same OSV format under its own identifiers, in three kinds: GitHub-reviewed; unreviewed, which are taken automatically from the NVD feed and "not checked for validity or completeness"; and malware, sourced from the npm security team and the OpenSSF malicious-packages repository [10]. Two further records cover the surrounding ground. abuse.ch, begun by one researcher in 2008 and licensed to Spamhaus Technology since 1 August 2022, runs free platforms for malware samples, malicious URLs, indicators of compromise and YARA rules [11]. Have I Been Pwned, run by Troy Hunt since 2013, lets a person check whether an account appeared in a data breach [12]. ## What changed in 2025 and 2026 **The CVE Program's funding.** On 15 April 2025 it was reported that MITRE's contract would expire the next day; CISA extended it by 11 months before it lapsed [4, secondary]. A CVE Foundation was formed with the stated aim of moving the programme "from a single-funding stream to a diversified funding model" [13]. On 23 September 2026 CISA published a whitepaper, "CVE Program: Establishing a Quality Era Framework," and stated that it "remains committed to leading, growing and sustaining the CVE Program into the foreseeable future" [14]. The programme continued to issue identifiers through 2026: advisories carrying CVE-2026 numbers were posted to Full Disclosure and oss-security in October 2026 [1][2]. **The NVD's scope.** On 15 April 2026 NIST moved the NVD to a risk-based operating model. Only prioritised CVEs receive enrichment; the rest are labelled "Lowest Priority - not scheduled for immediate enrichment." Backlogged CVEs published before 1 March 2026 were moved to a "Not Scheduled" status, and NIST no longer routinely issues its own severity score when the CNA has provided one. NIST's announcement states that CVE submissions "increased 263% between 2020 and 2025" and that it enriched nearly 42,000 CVEs in 2025 [5]. On 28 April 2026 NIST corrected about 4,500 CVSS v4.0 scores, 19% of those with such a score, which had been too high; on 17 June 2026 it added CISA's SSVC decision data and "affected" product information to its feeds and API [5]. The Cloud Security Alliance's reading of the April change is that full enrichment now covers an estimated 15 to 20% of CVE volume [15, secondary]. **New records in Europe.** ENISA's European Vulnerability Database was announced on 13 May 2025 under the NIS2 Directive; ENISA has been a CVE Numbering Authority since January 2024 [16, secondary]. On 13 October 2026 the database's website returned "Application Unavailable" [17]. The Global CVE Allocation System, announced on 16 April 2025, describes itself as community-led and aims to stay compatible with CVE through independent numbering authorities; its directory lists CVE as GNA 0 and CIRCL as GNA 1 [18]. ## Where the AI-specific risks are recorded The records above are organised around a product version and a fix. Three public records cover failures of AI systems, and none of them is organised that way. - **MITRE ATLAS** is "a public knowledge base of adversary TTPs targeting AI systems," at data version 2026.05. It lists LLM Prompt Injection as technique AML.T0051, with sub-techniques AML.T0051.000 Direct, in which "an adversary may inject prompts directly as a user of the LLM," and AML.T0051.001 Indirect, in which the prompt arrives through data the model ingests [19]. - **AVID**, the AI Vulnerability Database, run by the AI Risk and Vulnerability Alliance, is an open-source knowledge base of failure modes for AI systems, including "end-to-end AI applications and agents," built from reproducible evaluation evidence [20]. - The **AI Incident Database**, a project of the Responsible AI Collaborative, records real-world harms from AI systems under a deliberately broad definition, from an autonomous car killing a pedestrian to a wrongful arrest from face recognition; it began with more than 1,000 incident reports [21]. See [[Prompt Injection]] for why that risk has no version to patch. ## The same shape in other fields A public record with its own identifiers and a named keeper is not specific to security. The SEC's EDGAR holds "millions of company and individual filings," processes about 4,700 filings a day, and is free to the public [22]. Free Law Project's RECAP archive, searchable on CourtListener, holds "tens of millions" of federal court documents bought from PACER by its users and shared [23]. Wikidata records structured statements together with their sources under a public-domain licence, each item identified by a Q-number [24]; this vault's own check for deaths among the people quoted on [[Programmers Quotes]] reads from it. For the web, Google's Chrome User Experience Report holds real-user Core Web Vitals for eligible public sites and feeds Google Search's page-experience ranking factor [25], and HTTP Archive, part of the Internet Archive, keeps "a permanent repository of web performance information" [26]. ## Sources 1. [oss-security mailing list archive — Openwall](https://www.openwall.com/lists/oss-security/) *(primary)* 2. [Full Disclosure mailing list archive — SecLists](https://seclists.org/fulldisclosure/) *(primary)* 3. [About the Zero Day Initiative](https://www.zerodayinitiative.com/about) *(primary)* 4. [Common Vulnerabilities and Exposures — Wikipedia](https://en.wikipedia.org/wiki/Common_Vulnerabilities_and_Exposures) *(reputable secondary; the programme's own site could not be read by the tool)* 5. [National Vulnerability Database news — NIST](https://www.nist.gov/itl/nvd) *(primary: the 15 April 2026 operating-model announcement and the 28 April and 17 June 2026 updates)* 6. [Known Exploited Vulnerabilities Catalog — CISA](https://www.cisa.gov/known-exploited-vulnerabilities-catalog) *(primary)* 7. [Exploit Prediction Scoring System — FIRST](https://www.first.org/epss/) *(primary)* 8. [ATT&CK resources — MITRE](https://attack.mitre.org/resources/) *(primary)* 9. [OSV](https://osv.dev/) *(primary)* 10. [About the GitHub Advisory Database — GitHub Docs](https://docs.github.com/en/code-security/security-advisories/working-with-global-security-advisories-from-the-github-advisory-database/about-the-github-advisory-database) *(primary)* 11. [abuse.ch](https://abuse.ch/) *(primary)* 12. [About Have I Been Pwned](https://haveibeenpwned.com/About) *(primary)* 13. [CISA Statement on CVE — The CVE Foundation, 23 Apr 2025](https://www.thecvefoundation.org/newsroom/posts/2025-04-23-cisa-statement) *(primary for the Foundation's stated aim)* 14. [CISA Whitepaper Charts Path to Establishing and Maturing CVE Program Quality — CISA, 23 Sept 2026](https://www.cisa.gov/news-events/news/cisa-whitepaper-charts-path-establishing-and-maturing-cve-program-quality) *(primary: announces "CVE Program: Establishing a Quality Era Framework")* 15. [NVD Triage Overhaul: End of Universal CVE Enrichment — Cloud Security Alliance, Apr 2026](https://labs.cloudsecurityalliance.org/research/csa-research-note-nist-nvd-enrichment-overhaul-20260429-csa/) *(reputable secondary)* 16. [EU cybersecurity agency ENISA launches European Vulnerability Database — SecurityWeek, May 2025](https://www.securityweek.com/eu-cybersecurity-agency-enisa-launches-european-vulnerability-database/) *(reputable secondary)* 17. [European Vulnerability Database — ENISA](https://euvd.enisa.europa.eu/) *(primary; returned "Application Unavailable" on 13 Oct 2026)* 18. [GCVE — Global CVE Allocation System](https://gcve.eu/) *(primary)* 19. [ATLAS data — MITRE, technique AML.T0051](https://github.com/mitre-atlas/atlas-data) *(primary)* 20. [AI Vulnerability Database — AI Risk and Vulnerability Alliance](https://avidml.org/) *(primary)* 21. [About the AI Incident Database — Responsible AI Collaborative](https://incidentdatabase.ai/about/) *(primary)* 22. [About EDGAR — U.S. Securities and Exchange Commission](https://www.sec.gov/edgar/about) *(primary)* 23. [RECAP — Free Law Project](https://free.law/recap/) *(primary)* 24. [Wikidata: Introduction](https://www.wikidata.org/wiki/Wikidata:Introduction) *(primary)* 25. [Chrome UX Report — Chrome for Developers](https://developer.chrome.com/docs/crux) *(primary)* 26. [About HTTP Archive](https://httparchive.org/about) *(primary)* ## Related - [[AI Security for Users]] — the section overview - [[Exploit Database]] — the demonstration record, and the note this one accompanies - [[AI Supply Chain Risks]] — the package advisories above, seen from the user's side - [[Prompt Injection]] — the AI-specific risk, and why no version-bound record can hold it --- *Drafted with AI assistance from single-pass research; sources were cited by the tool and not opened by a person. Errors: see [[Corrections]].*