GitLab warns of critical RCE vulnerability in AI Gateway service
GitLab urges immediate patches for critical AI Gateway RCE CVE-2026-90970 on self-hosted instances.
GitLab warned that CVE-2026-90970, an improper neutralization flaw in the AI Gateway, lets an authenticated user with Duo Agent Platform access escape the prompt-template sandbox through a crafted flow configuration and run arbitrary commands. Fixes are in versions 19.2.4, 19.3.2, and 19.4.1 for self-hosted gateways; GitLab-hosted gateways are already protected. The report also notes last month's unauthenticated path traversal CVE-2026-85706 in GitLab CE and EE, which CISA added to its exploited-vulnerability catalog under BOD 26-04.
- CVE-2026-90970 allows authenticated command execution on the AI Gateway
- Self-hosted installs should move to 19.2.4, 19.3.2, or 19.4.1
- GitLab-hosted AI Gateway customers are already protected
- CISA lists path traversal CVE-2026-85706 as actively exploited
- GitLab cites over 30 million users and broad Fortune 100 use
- Vendors
- GitLab
- Organizations
- CISA
- Countries
- United States
Vulnerabilities mentioned
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
| CVE-2026-85706 | Unauthenticated Path Traversal Arbitrary File Read in GitLab CE/EE | 10.0 | 97% | KEVPoC ×13 | KEV |
| CVE-2026-90970 | Prompt sandbox escape to RCE in GitLab AI Gateway | 9.9 | 17% | PoC |
From the original article
2 min read at bleepingcomputer.com
GitLab warned customers today to immediately patch a critical AI Gateway vulnerability that could let attackers run arbitrary commands on vulnerable instances. [...]
By Sergiu Gatlan · bleepingcomputer.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.