GitLab is addressing CVE-2026-90970, a critical vulnerability in its self-hosted AI Gateway that could allow arbitrary command execution. The company disclosed the flaw on October 2 and urged customers to apply the necessary patches immediately. The vulnerability has a CVSS score of 9.9 out of 10. An authenticated user could exploit it to escape the prompt template sandbox and run commands on the gateway. Users of GitLab’s cloud-hosted AI Gateway are already protected.
Separately, CVE-2026-85706, a critical path-traversal vulnerability, was reportedly under active exploitation as of October 3-4, 2026. The flaw could allow an unauthenticated remote attacker to read arbitrary files from a GitLab server. Exploitation requires only that one public project exist.
The vulnerabilities pose significant risks for users of GitLab’s widely adopted platform, which numerous major corporations use.