Critical GitLab File-Read Flaw Exploited Within a Day of Patch Disclosure

Threat actors began probing a critical GitLab vulnerability only one day after the company announced security updates, according to researchers monitoring exploitation activity.

The flaw, tracked as CVE-2026-85706, allows an unauthenticated attacker to read arbitrary files from a self-hosted GitLab server. Exploitation requires a single crafted HTTP request, making exposed and unpatched installations practical targets for automated scanning.

CVE-2026-85706 has a CVSS score of 10 and is already in the US Cybersecurity and Infrastructure Security Agency’s Known Exploited Vulnerabilities catalog. CISA added it on September 11, 2026, and gave federal agencies until September 14, 2026, to remediate affected systems.

The associated GitLab release fixes 17 other vulnerabilities, including a second critical defect and six high-severity issues. Administrators should therefore deploy the complete update rather than treating the incident as a narrowly isolated path-traversal bug.

Unauthenticated path traversal reaches server files

CVE-2026-85706 is a path-traversal vulnerability in the GitLab repository commits API. Improper confinement of file paths, combined with missing authentication enforcement, permits requests to escape the intended path and access files elsewhere on the server.

No valid GitLab account is required. The attack can be delivered through one HTTP request.

Its CVSS vector is CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:N. This reflects a remotely exploitable vulnerability with low attack complexity, no privileges or user interaction required, and a high confidentiality impact.

The vulnerability does not directly receive an availability impact under that scoring vector. However, arbitrary-file access can expose information that enables subsequent attacks. Files on a GitLab host may contain application configuration, credentials, access tokens, source code, integration secrets, or details about connected infrastructure.

Those consequences depend on which files are accessible and how each installation is configured. Public reporting has not identified the specific files retrieved during the observed exploitation attempts, nor has it disclosed confirmed downstream compromises attributable to those attempts.

WatchTowr observed in-the-wild attempts one day after GitLab announced its patches and expects broader exploitation. The short interval leaves little separation between vulnerability disclosure, reverse engineering, scanning, and attempted compromise.

Exact GitLab CE and EE versions affected

The vulnerability affects self-hosted GitLab Community Edition and Enterprise Edition installations in these version ranges:

• GitLab CE and EE from 18.7 before 19.1.8
• GitLab CE and EE from 19.2 before 19.2.6
• GitLab CE and EE from 19.3 before 19.3.2

Administrators should upgrade to the fixed release appropriate for their installed branch:

• 19.1.8 or later
• 19.2.6 or later
• 19.3.2 or later

Organizations on earlier or separately maintained branches should obtain the corresponding vendor-supported fixed release. No cloud-specific affected versions or mitigation instructions have been disclosed.

Internet-facing self-hosted instances face the most immediate scanning risk, but internal exposure should not be ignored. An attacker who already has access to a corporate network could target an internally reachable GitLab server using the same unauthenticated request.

GitLab systems also occupy a sensitive position within development environments. They commonly connect to runners, deployment platforms, registries, identity providers, ticketing systems, and cloud infrastructure. Access to files on the host could therefore expose credentials with reach beyond GitLab itself.

Log searches can reveal attempted exploitation

GitLab administrators should review HTTP logs for POST requests sent to the repository commits API:

/api/v4/projects/{id}/repository/commits/

Requests containing a file.path parameter warrant investigation because they may represent exploitation attempts against CVE-2026-85706.

That pattern alone should not automatically be treated as proof of compromise. Investigators should correlate matching requests with source IP addresses, timestamps, response codes, unusual file access, account activity, and subsequent changes to repositories or pipelines.

A response should also examine:

• Repository commits and modifications around the suspicious request
• Access to CI/CD variables and project secrets
• Pipeline definitions and recently triggered jobs
• Runner activity and unexpected build operations
• Deploy keys, personal or project access tokens, and integrations
• Changes involving users, permissions, webhooks, or protected branches
• Newly generated or altered build artifacts
• Authentication using credentials stored on the GitLab host
• Activity in downstream deployment and package environments

Potentially exposed credentials and secrets should be rotated. Teams should validate pipeline definitions and rebuild sensitive artifacts from trusted source states when compromise cannot be excluded.

Patching closes the known flaw but does not remove an attacker who already obtained credentials or modified development assets. Systems showing suspicious requests require forensic review in addition to upgrading.

CISA sets a September 14 federal deadline

CISA placed CVE-2026-85706 in its KEV catalog on September 11, 2026. Inclusion confirms that exploitation is occurring outside controlled research environments and changes the issue from a theoretical exposure into an active operational priority.

US federal civilian agencies must complete remediation by September 14, 2026.

CISA requires affected organizations to apply vendor mitigations while complying with BOD 26-04 guidance for risk-based prioritization of security updates. Agencies must also follow CISA’s Forensics Triage Requirements.

For cloud services, organizations are directed to follow the applicable BOD 26-04 guidance or discontinue use when effective mitigations are unavailable. Each stakeholder remains responsible for evaluating an asset’s internet exposure and applying the relevant patching requirements.

No workaround that provides equivalent protection has been identified. Immediate installation of the fixed GitLab release is therefore the primary defensive action.

The update closes another critical GitLab weakness

The same security release fixes CVE-2026-87719, an insecure-deserialization vulnerability in GitLab’s GraphQL subscription serializer.

It carries a CVSS score of 9.9 and the vector CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H. Unlike CVE-2026-85706, exploitation requires low-level privileges, but the potential effects span confidentiality, integrity, and availability.

An attacker could gain access to Advanced Search instance configurations and sensitive credentials. GitLab CE and EE versions 19.1.8, 19.2.6, and 19.3.2 resolve this issue.

The release additionally addresses six high-severity vulnerabilities capable of enabling remote code execution, access to protected CI/CD variables, cross-site scripting, or denial of service. Details about the remaining fixed defects and all their affected version ranges are not available here.

Deploying the full release matters. A system patched only for the path-traversal route could remain exposed through one of the other corrected components.

Rapid exploitation follows an earlier GitLab case

The latest activity is not GitLab’s only recent encounter with near-immediate exploitation. Researchers previously observed rapid attacks targeting CVE-2026-19478, a critical GraphQL code-injection vulnerability.

CVE-2026-19478 has a CVSS score of 9.4 and is classified as CWE-94. Under certain conditions, an unauthenticated attacker could abuse a GraphQL directive to remotely modify or delete public projects and user data.

Its affected ranges were:

• GitLab CE and EE from 18.2 before 18.11.11
• From 19.0 before 19.0.8
• From 19.1 before 19.1.6
• From 19.2 before 19.2.4

The recurrence shows how quickly attackers can operationalize GitLab flaws, particularly those reachable without authentication. Development platforms hold valuable code, secrets, and automation controls in one location.

Administrators should patch CVE-2026-85706 immediately, preserve relevant logs, and investigate suspicious repository-commits API traffic. Any evidence of file access should trigger credential rotation and a broader review of pipelines, runners, integrations, artifacts, and downstream deployments.