CVE-2026-45387: Open WebUI System Prompt Exposure via Shared Model Permissions

# CVE-2026-45387: Open WebUI System Prompt Exposure via Shared Model Permissions

https://gcdradar.vercel.app/incidents/cve-2026-45387-open-webui-system-prompt-exposure-via-shared · AIR-2026-0698 · closed 2026-05-14

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2026-05-14
- AI role: Attack on AI
- Products: Open WebUI
- CVE ids: CVE-2026-45387
- CVSS: 4.3

## Summary

A vulnerability in Open WebUI permits users with read access to shared models to view confidential system prompts. The issue impacts confidentiality across shared workspace configurations.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-05-15: CVE-2026-45387: Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. (https://nvd.nist.gov/vuln/detail/CVE-2026-45387)
- 2026-05-14: CVE-2026-45387: Open WebUI: Sharing models for others to use (read permission) also exposes model details (system prompt leakage) (https://github.com/advisories/GHSA-h2cw-7qw9-56xr)

## Sources

- 2026-05-14 GitHub Advisory Database: CVE-2026-45387: Open WebUI: Sharing models for others to use (read permission) also exposes model details (system prompt leakage) https://github.com/advisories/GHSA-h2cw-7qw9-56xr
- 2026-05-15 NVD: CVE-2026-45387: Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. https://nvd.nist.gov/vuln/detail/CVE-2026-45387

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.