GitLab warns of critical 9.9 AI Gateway flaw allowing command execution

2026-10-04 · Cybersecurity

GitLab patched a critical vulnerability in its AI Gateway, the service that connects a GitLab instance to AI models, that could let a logged-in user with Duo Agent Platform access run commands on vulnerable servers. Only organizations hosting their own gateway need to act, and GitLab urged immediate patching.

Why it matters：The flaw affects organizations hosting their own GitLab AI Gateway and highlights the security risks of services that connect code repositories to external AI models.

Key points

• The flaw is rated 9.9 and allows command execution on self-hosted servers
• Only organizations that host their own AI Gateway are affected
• Exploitation requires a logged-in user with Duo Agent Platform access

Sources (original reporting)

• The Hacker News
• BleepingComputer

More stories from this day

• China-linked Warlock ransomware exploits SharePoint flaws in critical-infrastructure attacks
• China-nexus Antino backdoor targets Asian governments via Outlook and OneDrive
• Technical University of Denmark breach may expose data of 200,000 people
• ShinyHunters suspect detained in Jordan, reportedly cooperating with FBI
• Nvidia and SoftBank close final $20B investment in OpenAI round
• FlyDubai co-pilot's axe attack ruled attempted terrorism