GitLab Patches Critical AI Gateway Vulnerability Allowing Command Execution

GitLab has issued an advisory regarding a critical security flaw in its AI Gateway service, identified as CVE-2026-44951. The vulnerability stems from insufficient access controls within the gateway component, which connects self-hosted GitLab instances to external AI models. A logged-in user possessing Duo Agent Platform access could potentially exploit this weakness to execute arbitrary commands on the underlying server hosting the gateway.

The AI Gateway service is designed to facilitate integration between GitLab and various AI model providers. However, the scope of impact is limited to organizations that self-host the AI Gateway component. GitLab-managed cloud services are not affected by this specific vulnerability.

The flaw was patched in the following releases: gateway version 19.2.4, 19.3.2, and 19.4.1. GitLab has not confirmed whether the vulnerability is being actively exploited in the wild, though the advisory notes the potential for command execution under the specified conditions. Security researchers and IT teams are advised to verify their current gateway version and apply the necessary updates to mitigate the risk.