Servers and applications
CVE-2026-44563
Ollama Model Access Control Bypass via /api/generate, /api/embed, /api/embeddings, and /api/show
- Severity
- Medium 5.4
- Published
- 5 May 2026
- Last changed
- 11 Nov 2026
- Added to UpHeron
- 1 Nov 2026
- Fixed in
- 0.9.0
- Exploit chance
- 0.2% in the next 30 daysHigher than 15% of CVEs. EPSS by FIRST, 19 Nov 2026
- Weakness
- Missing authorization (CWE-862)
- Attack
- Over the network · Needs a user account · No user action needed
- Attacker can
- Read some data · Slow it down
- CVSS vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:L- Also known as
- GHSA-rcvp-6fgw-c7fh
- Sources
- NVD, Project advisories (GitHub)
Description
Four Ollama proxy endpoints accept any model name from the user and forward the request to the Ollama backend without checking whether the user is authorized to access that model. These endpoints only require get_verified_user (any authenticated non-pending user) and validate that the model exists in the full unfiltered model list, but never check AccessGrants.has_access().
This is in direct contrast with the /ollama/api/chat endpoint (line 1101-1122) which correctly validates model access grants and returns 403 for unauthorized users:
- Admin configures model access control, restricting llama3:70b to the "ML Engineers" group. Regular user Alice is only authorized for llama3:8b.
- Alice knows the restricted model name (model names are predictable — llama3:70b, mistral:latest, etc.).
- Alice calls the unprotected endpoints directly:
- All requests succeed and are proxied to Ollama without any access control check.
Affected software
| Software | Affected versions |
|---|---|
Open WebUIServers and applications · cpe:a:openwebui:open_webui |
|
Open WebUIServers and applications · github:open-webui/open-webui |
|
About Open WebUI
self-hosted user interface for artificial intelligence models
- Weekly downloads (PyPI)
- 355,496PyPI Stats, 15 Nov 2026
- Docker Hub pulls
- 990,663Docker Hub, 15 Nov 2026
- GitHub stars
- 158,684GitHub, 15 Nov 2026
- Forks
- 23,075GitHub, 15 Nov 2026
- Latest release
- v0.13.3GitHub, 15 Nov 2026
- Released
- 8 Nov 2026GitHub
- License
- multi-licensingWikidata
- First published on GitHub
- 6 Oct 2023GitHub
Website Source code Docker image Package
Vulnerabilities in Open WebUI
- In our list
- 177
- Published in the last 12 months
- 162
- Critical, last 12 months
- 5
- High, last 12 months
- 67
- Known exploited
- 0
- GHSA-3vvh-g22g-jq6rChannel members can post messages shown as a webhook via client-supplied message metadataMedium 4.3 18 Oct 2026
- CVE-2026-109572Terminal users can manage Terminals policies and other users' terminals via the terminal proxyHigh 7.2 28 Sept 2026
- CVE-2026-109292Other users' session tokens disclosed via OAuth-mode connections, tool servers and terminalsHigh 7.6 27 Sept 2026
- CVE-2026-109297Instance-wide stall via automation recurrence rules that evade sub-daily frequency detectionHigh 7.1 27 Sept 2026
- CVE-2026-109288Viewer session token theft via javascript: URLs in chat message linksHigh 7.2 27 Sept 2026
- CVE-2026-109294Any authenticated user can exhaust server memory via a compressed remote image in a chat messageHigh 7.1 27 Sept 2026
References
- NVD https://nvd.nist.gov/vuln/detail/CVE-2026-44563
- CVE.org https://www.cve.org/CVERecord?id=CVE-2026-44563
- GitHub Advisory Database https://github.com/advisories/GHSA-rcvp-6fgw-c7fh
- github.com https://github.com/open-webui/open-webui/security/advisories/GHSA-rcvp-6fgw-c7fh