Improper neutralization of input used for LLM prompting in GitLab Enterprise Edition - CVE-2026-96893
Published: October 13, 2026
Vulnerability identifier: #VU154154
CSH Severity: Low
CVSS v4: 8.4 [CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:A/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N]
CVE-ID: CVE-2026-96893
CWE-ID:
Exploitation vector: Remote access
Exploit availability:
No public exploit available
Vulnerability details
The vulnerability allows a remote user to potentially cause unintended command execution attributed to other users.
The vulnerability exists due to improper neutralization of input used for LLM prompting in GitLab Duo Agent Platform when constructing AI conversation context. A remote user can inject specially crafted content into AI agent instructions to potentially cause unintended command execution attributed to other users.
User interaction is required.
Affected software
GitLab Enterprise Edition
How to mitigate CVE-2026-96893
Install security update from vendor's website.
GitLab Enterprise Edition - addressed in versions 19.2.8, 19.3.4, 19.4.2