GitLab AI Gateway Flaw Scores 9.9 CVSS

On October 3, 2026, DEV Community published a report saying GitLab had issued emergency security releases to fix a critical vulnerability affecting self-hosted deployments of its AI Gateway. The vulnerability is tracked as CVE-2026-90970 and carries a CVSS v3.1 score of 9.9, which the report flags in its headline as a command execution risk.

According to the report, the flaw stems from improper neutralization of special elements used in template engines, a weakness it maps to CWE-1336. As a result, administrators or other authorized users who customize prompt flow templates face a command execution risk. The writeup, which includes a technical vulnerability analysis section, describes the exposure as specific to self-hosted installations of AI Gateway, and it characterizes GitLab's response as an emergency security release.

The recommendation given to readers is straightforward: teams running a self-hosted AI Gateway should upgrade as soon as possible. The report does not specify which versions contain the fix, does not state whether the vulnerability has been exploited, and does not say who discovered or reported it. It gives no date for when the vulnerable code was introduced or when the fix shipped, and it offers no statement from GitLab beyond the fact that emergency releases were made available. As of this report, that is the extent of what is known: a 9.9-rated command execution flaw in self-hosted AI Gateway, a CWE-1336 root cause tied to prompt flow templates, and a call to patch.