Free AI Safety Workshop — in collaboration with explainx.aiSave your seat
AgentbeamBeta
AI agent vulnerability · Open WebUI

CVE-2026-45347: Open WebUI vulnerable to blind server side request forgery (SSRF) via the PDF generate function

mediumCVSS 3.1 4.3GHSA-f776-fp4w-266c

Summary Blind server side request forgery (SSRF) via the PDF generate function. The finding resulted from a penetration test for a customer. It is suspected that the root cause of the issue lies within the core of Open WebUI, which is why it is being reported as a security issue here. Tested on Open WebUI 0.5.4. Details In the PDF export, user inputs are interpreted as HTML and embedded into the PDF.

Published May 14, 2026 · Updated May 19, 2026

Affected packages

open-webui

Ecosystem
PyPI
Affected
< 0.5.11
Fixed in
0.5.11
Check
pip show open-webui

Advisory details

From the GitHub Advisory Database (CC BY 4.0).

Summary

Blind server side request forgery (SSRF) via the PDF generate function. The finding resulted from a penetration test for a customer. It is suspected that the root cause of the issue lies within the core of Open WebUI, which is why it is being reported as a security issue here. Tested on Open WebUI 0.5.4.

Details

In the PDF export, user inputs are interpreted as HTML and embedded into the PDF. According to tests, scripts and some potentially dangerous tags (iFrame, Object, etc.) are blocked, preventing server-side content from being read through this vulnerability. However, an image tag can be used to force a server-side request (SSRF), as shown in the following below.

PoC

Start a chat and export the PDF: grafik

Intercept the request and insert an ` tag into the title`:

POST /api/v1/utils/pdf HTTP/2
Host: domain.local
//Some headers removed
Content-Type: application/json
Content-Length: 541
Te: trailers

{"title":"","messages":[{"id":"81f24589-384d-431c-a26c-5cd3382ac941","parentId":null,"childrenIds":["0c1a3ee1-6350-4bb4-b95e-fc2341c47e8e"],"role":"user","content":"hallo","timestamp":1736932102,"models":["gpt-4o-POC"]},{"parentId":"81f24589-384d-431c-a26c-5cd3382ac941","id":"0c1a3ee1-6350-4bb4-b95e-fc2341c47e8e","childrenIds":[],"role":"assistant","content":"Hallo! Wie kann ich Ihnen helfen?","model":"gpt-4o-POC","modelName":"gpt-4o-POC","modelIdx":0,"userContext":null,"timestamp":1736932103,"done":true}]}

A HTTPS callback was received at https://d5jok0s7ghl1p77v5brlqlxwmnsega4z.oastify.com.

Impact

A user can force server-side GET requests. During the available testing time, no method was found to read the responses (Blind SSRF). Nonetheless, this should be prevented, as an attacker could enumerate internal assets through response delays and trigger arbitrary GET requests.

Resolution

Fixed in commit 167c8bf00, first released in v0.5.11 (2025-02). The fix wraps every user-controllable field that flows into the PDF HTML template (title, content, role, model, formatted date) in html.escape() before the template f-string is fed to fpdf2.write_html(). The PoC payload ` is escaped to &lt;img src=&#x27;...&#x27; /&gt; and rendered as literal text by fpdf2, with no HTML parsing and no outbound request. Users on >= 0.5.11` are not affected.

Credited to simioni87.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. No analysis has been written for this entry yet. Methodology

More Open WebUI vulnerabilities

All Open WebUI entries