Summary
- CVE-2026-85706 is a CVSS 10.0 path-traversal flaw that can allow unauthenticated arbitrary-file access on affected self-managed GitLab systems.
- CISA added the vulnerability to its Known Exploited Vulnerabilities catalogue on 11 September, establishing real-world exploitation.
- watchTowr separately observed behavioural probes after reproducing the flaw, although public sources do not yet establish the scale or victims of exploitation.
A maximum-severity vulnerability in GitLab has been added to the US government’s Known Exploited Vulnerabilities catalogue, moving the evidence around the flaw beyond rapid reconnaissance to confirmed exploitation in the wild.
CVE-2026-85706 is a path-traversal vulnerability in the repository commits API affecting GitLab Community Edition and Enterprise Edition. GitLab assigned it the maximum CVSS score of 10.0.
The company says that, under certain conditions, an unauthenticated user can read arbitrary files from an affected GitLab server because of improper path confinement and missing authentication enforcement.
Affected releases include versions from 18.7 before 19.1.8, the 19.2 branch before 19.2.6, and the 19.3 branch before 19.3.2. GitLab.com was already running the patched version when the advisory was issued, while GitLab Dedicated customers do not need to take action for the vulnerability.
The US Cybersecurity and Infrastructure Security Agency added CVE-2026-85706 to its Known Exploited Vulnerabilities catalogue on 11 September. CISA’s listing establishes that exploitation has occurred but does not identify victims, the actors involved, or the scale of the activity.
Security company watchTowr had separately reproduced the issue and observed behavioural probes targeting its global honeypot infrastructure. The researchers assessed that attackers had successfully reverse engineered the vulnerability soon after disclosure.
The two evidence points should not be conflated. Honeypot probes show active reconnaissance and attempts to identify vulnerable systems; CISA’s catalogue entry establishes a separate threshold by classifying the vulnerability as known to have been exploited.
The affected platform makes that development particularly consequential. Self-managed GitLab environments can contain source code, configuration files, continuous-integration settings, deployment workflows, credentials, and other material used to move software from development into production.
An arbitrary-file-read vulnerability does not automatically translate into complete compromise of every affected instance. The practical outcome depends on which files are reachable, how an installation is configured, what credentials or secrets are present, and what other systems trust the GitLab environment.
Development platforms nevertheless concentrate access that can reach far beyond the host itself. Pipeline credentials may connect to cloud infrastructure, package registries, container platforms, and production environments, while repository data can expose proprietary code and security-relevant configuration.
The incident also illustrates the shrinking interval between vulnerability disclosure and attack activity. Researchers and hostile operators can compare patches, reproduce flaws, and automate scanning quickly enough that urgent releases may begin attracting external traffic before many organisations have completed normal change-control processes.
A separate critical GitLab vulnerability patched in August affected the platform’s GraphQL layer and should not be confused with CVE-2026-85706. The latest flaw is distinct, but the repeated urgent releases reinforce the concentration of risk around self-managed development infrastructure.
GitLab also corrected two other serious issues in the same September release, including an insecure-deserialisation vulnerability in Enterprise Edition and a buffer overflow associated with Advanced Search indexing. Neither should be assumed to share the exploitation status of CVE-2026-85706 without separate evidence.
The current evidence is therefore strong but bounded: the path-traversal flaw is critical, fixes are available, hostile probing followed disclosure quickly, and CISA now classifies it as exploited. What remains unknown is how widely successful attacks have occurred and what attackers have obtained from compromised systems.
