Skip to main content

Vulners

ProductVulners
Records still shown as active9, every withdrawn identifier against Open WebUI
First contacted2026-07-23
Channels triedsupport@
StatusReplied the same day on one record and opened an internal task. Not yet reflected, and the wider problem is unaddressed.
note

Vulners replied the same day, identified the cause and opened an internal task. The findings on this page are not a complaint about responsiveness.

Records​

Every CVE withdrawn against Open WebUI still has a live Vulners page. Each identifier below is in the REJECTED state at cve.org and at NVD:

IdentifierWithdrawn byRejected sinceStill live on VulnersOur assessment
CVE-2024-7033huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2024-7034huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2024-7038huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2024-7039huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2024-7040huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2024-7959huntr / Protect AI2026-07-16vulners.comDisposition
CVE-2025-15603VulDB2026-06-18vulners.comDisposition
CVE-2025-29446MITRE2026-06-29vulners.comDisposition
CVE-2025-63391MITRE2026-06-29vulners.comDisposition

The withdrawal date is on the record, and the record is marked current​

The Vulners entry for CVE-2025-63391 is stamped 29 June 2026 and labelled Current. That is the same date MITRE withdrew the identifier. The entry was therefore last touched on the day of the withdrawal, and it still presents the finding as live, at 6.9, Medium risk, on Vulners' own AI Score rather than the record's CVSS of 7.5, with an EPSS value and a set of topic tags reading "authentication bypass", "unauthenticated attackers" and "sensitive configuration".

Whatever ran on that date did not carry the rejection through.

Generated text that recommends upgrading​

The description shown by default is machine-generated and marked "AI enhanced". For CVE-2025-63391 it states that the endpoint allows unauthenticated remote attackers to access sensitive configuration data, cites other databases as corroboration, and closes with remediation guidance: that affected organisations "should upgrade to a patched version of Open-WebUI or apply vendor-provided mitigations".

There is no patched version, because there was never a defect. An operator following that guidance would go looking for an upgrade that does not exist, to remediate a finding the issuing CNA has withdrawn. The guidance is also stated more confidently than the original report ever put it.

The record is propagated into package advisories against other people's software​

This is where the cost lands on people who had no part in it. Under its vulnersOsv source, Vulners maps the withdrawn identifiers onto downstream Python packages that depend on Open WebUI:

  • For CVE-2025-63391: hubzoid (>=0.2.2, <=0.6.0) and openwebui-token-tracking (>=0.1.7, <=0.1.10) and one further package, listed as potentially affected via open-webui (>=0.6.0, <=0.8.8).
  • For CVE-2025-29446: openwebui-token-tracking (=0.1.7), listed as potentially affected via open-webui (=0.6.0).

These are not our packages and they are not our users. They are third-party projects, now carrying a dependency-level advisory for an identifier that no longer exists, generated from a record its own CNA withdrew. Their maintainers did nothing to earn it, and any scanner consuming this mapping inherits the advisory.

A single withdrawn record does not stay in one database. The same entry lists eleven related sources carrying it onward, including AttackerKB, CIRCL, CNNVD, EUVD, Positive Technologies, Red Hat and Snyk. That is the whole problem in one screen: rejection at cve.org is a single event, and the copies do not follow.

CVE-2025-15603 in detail​

Authoritative stateREJECTED at cve.org and NVD since 2026-06-18
Withdrawn byVulDB, the issuing CNA, as a false positive
CNA rating before withdrawalLow (CVSS v2 2.6, v3 3.7, v4 2.9)
What Vulners displaysAn active vulnerability in open-webui, with a machine-generated Vulners AI Score of 5.3 (Medium)
Our assessmentCVE-2025-15603

The underlying data is correct here: Vulners confirmed the record carries vulnStatus: "Rejected" in the API. The web interface does not surface that field, so the entry reads as live to anyone browsing it.

The score is a separate point. The AI Score of 5.3 (Medium) sits above the CNA's own rating of Low, for a record the CNA has withdrawn. A machine-generated score that exceeds the issuing CNA's assessment should be presented as the database's own opinion, not as the record.

Contact log​

DateChannelOutcome
2026-07-23support@Replied the same day. Confirmed the API carries vulnStatus: "Rejected", acknowledged the web interface does not reflect it, and opened an internal task to surface the status correctly.
2026-08-08support@Follow-up asking for an update. The entry still does not show the rejected state or the updated record text.

As of 2026-08-08 the fix has not shipped, and the eight other withdrawn identifiers listed above have not been raised individually.

See also​

This content is for informational purposes only and does not constitute a warranty, guarantee, or contractual commitment. Open WebUI is provided "as is." See your license for applicable terms.