GitLab Patches Critical AI Gateway Flaw Allowing Command Execution

· Original article ↗

Summary

CVE-2026-90970 could let authenticated users with basic privileges and Duo Agent Platform access execute commands on vulnerable self-hosted GitLab AI Gateway instances. GitLab released fixes; hosted customers are already protected.

Key points

  • CVE-2026-90970 could allow arbitrary command execution by escaping the prompt-template sandbox through a specially crafted flow configuration.
  • Exploitation requires authentication, basic privileges, and Duo Agent Platform access.
  • The flaw affects self-hosted GitLab AI Gateway deployments.
  • GitLab fixed the vulnerability in versions 19.2.4, 19.3.2, and 19.4.1 and urges affected customers to upgrade immediately.
  • Customers using GitLab-hosted AI Gateway instances are already protected and do not need to take action.
  • GitLab separately patched CVE-2026-85706, a path traversal flaw that CISA added to its actively exploited vulnerabilities list.

Article Details

Vulnerability Types
  • Improper neutralization
  • Sandbox escape leading to arbitrary command execution
  • Path traversal
Severity
Critical
Exploitation Status
reported
Exploit Availability
unknown
Patch Status
available

CVE

Vendors

Products

Related Articles