CVE-2026-129253
GitLab AI workflow tokens could bypass administrator tool governance
GitLab · GitLab Duo AI tool governance
First documented disclosure · Review checkpoint
Summary
A developer could bypass administrator-defined AI tool restrictions in a governed namespace.
Severity and exploitation evidence
Medium · 4.3/10 · CVSS 3.1. Severity source.
Exploitation not established: Malicious exploitation status remains unverified in the reviewed sources.
Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.
How the weakness works
Authorization tokens were insufficiently bound to their intended workflow context.
Conditions required
An authenticated user has developer-role permissions in the affected namespace.
Potential impact
Agent tool access can exceed administrator-defined governance controls.
Affected versions and configurations
19.1 through versions before 19.3.9, 19.4 before 19.4.8, and 19.5 before 19.5.6.
Fix and mitigation status
19.3.9, 19.4.8 and 19.5.6.
Practical next steps
Upgrade to a corrected release in the applicable maintenance branch.
Operating lesson
Bind tool authorization to the workflow execution context.
Evidence and limits
Original disclosure 2026-11-16. The CVE Program published this record 2026-11-17. CERT-FR’s November 17 GitLab bulletin also lists this identifier.
Sources
- GitLab CVE authority record
Primary source ·
- GitLab Duo AI tool governance original advisory
Primary source ·
- CERT-FR GitLab advisory
Primary source ·
- ENISA EUVD-2026-126838
Primary source ·