GitLab has patched a maximum-severity vulnerability that could have enabled attackers without any credentials to run arbitrary code on its servers. The company released fixes Monday for the critical flaw alongside five other high-severity bugs affecting its source code management and CI/CD platform. The vulnerabilities threatened pipeline security, repository access controls, and AI agent instruction handling across both Community and Enterprise editions.
The critical flaw, designated CVE-2026-97229, stems from an integer overflow problem in the platform's CI/CD regular expression compiler and carries the maximum possible CVSS score of 10.0. GitLab described the issue as allowing unauthenticated attackers to execute arbitrary code by submitting a specially crafted regular expression in a CI/CD configuration under certain conditions where insufficient validation occurred. The vulnerability affects Community Edition and Enterprise Edition versions from 19.2 before 19.2.8, 19.3 before 19.3.5, and 19.4 before 19.4.2, with fixes now available in those respective patched releases. The company's HackerOne bug bounty program received the report from researcher Johan Carlsson. GitLab.com already runs the patched version, while GitLab Dedicated customers don't need to take action, but administrators of self-managed installations running affected versions face urgent pressure to upgrade immediately.
The five high-severity vulnerabilities address a range of attack vectors across the platform. CVE-2026-92883, scoring 8.7 on the CVSS scale, could permit an authenticated user to run arbitrary JavaScript in other users' browser sessions through a specially crafted repository due to improper sanitization of HTML attributes in syntax-highlighted output. Two pipeline security flaws followed: CVE-2026-16301 with a CVSS score of 8.5 could have let users holding the Developer role run CI/CD pipelines on protected branches without proper authorization, while CVE-2026-14814 scoring 8.2 potentially allowed attackers to poison cached responses for raw repository files, causing targeted pipelines to execute malicious content and gain access to protected environment variables. The remaining issues tackle scheduled pipeline execution and GitLab Duo functionality. CVE-2026-81839, rated 7.7, could have permitted authenticated users to execute arbitrary pipelines in a transferred project using the new group's privileges. CVE-2026-96893, scoring 7.3, could have enabled an authenticated user to inject crafted content into AI agent instructions, potentially triggering unintended command execution attributed to other users.
The report states that GitLab plans to make vulnerability issue details public 90 days after the release in which they were patched. According to the company, administrators should prioritize the applicable security updates rather than wait for further technical disclosure. The release also addresses several low-to-medium severity vulnerabilities affecting GitLab Pages, Kubernetes proxying, GraphQL APIs, merge requests, custom roles, the package registry, and repository and project import functions. The critical regex flaw's 10.0 score reflects potential impacts on confidentiality, integrity, and availability across security boundaries, with remote code execution possible without any authentication requirement. Organizations running self-managed GitLab installations face a narrow window to patch before technical details become public and exploitation attempts likely follow. The breadth of vulnerabilities fixed in this release—from cross-site scripting and pipeline permission bypasses to AI agent instruction injection—illustrates the expanding attack surface as development platforms integrate more sophisticated automation and artificial intelligence features.