Critical Sandbox Bypass Flaw in GitLab AI Gateway Enables Remote Command Execution
GitLab has issued an urgent security update after a critical vulnerability was discovered in the GitLab AI Gateway. The flaw, identified as CVE-2026-90970, allows authenticated users to bypass sandbox protections and execute arbitrary commands on affected systems.
The issue exists in the prompt template processing logic used by custom flows within the Duo Agent Platform. Under certain conditions, an attacker with valid access can submit a specially crafted flow configuration that escapes the sandbox environment and runs operating system commands.
According to the advisory, the vulnerability received a CVSS v3.1 base score of 9.9 and is rated Critical. GitLab has published fixed releases for self-hosted environments: GitLab AI Gateway 19.4.1, 19.3.2, and 19.2.4. The company states that its own hosted AI Gateway service has already been patched.
Administrators of self-hosted GitLab instances are strongly advised to apply the updates without delay to prevent potential exploitation.
Related articles
WebPros Releases Critical Patches for Three Vulnerabilities in cPanel & WHM
WebPros International has published security updates addressing three critical vulnerabilities in its cPanel & WHM hosting management platform. All three issues received the highest severity rating of Critical. The flaws include CVE-2026-93698, an input validation weakness in the Multilang adminbin component that could allow root-level operating system command execution. Two additional stored cross-site scripting vulnerabilities were also fixed, one of which is CVE-2026-93697 affecting the Mass Modify Accounts interface in WHM. Successful exploitation of the XSS flaws could let low-privileged accounts hijack administrator sessions. The updates were made available on September 29, 2026, and carry CVSS v3.0 base scores reaching 9.9.
Top Cybersecurity Stories: SharePoint Exploits Warned by US Authorities, Citrix and WordPress Flaws Lead Weekly Rankings
Security NEXT has published its weekly ranking of the most viewed articles from September 27 to October 3, 2026, highlighting critical vulnerability disclosures and confirmed exploitation cases. US authorities issued warnings about active exploitation of five vulnerabilities affecting SharePoint and WordPress. Citrix NetScaler received multiple vulnerability advisories with two flaws already confirmed as exploited in the wild. Apple released iOS 26.7.1 to address vulnerabilities potentially used in targeted attacks against specific individuals. Other notable incidents include a personal data breach at Times Car car-sharing service and a ransomware attack impacting Keio Electric Railway operations.
Google Releases Chrome Update Fixing 11 Vulnerabilities Including Critical WebGL Flaw
Google has issued an update for its Chrome browser that addresses 11 security vulnerabilities across Windows, macOS, and Linux platforms. The release includes Chrome 154.0.8037.98 and 154.0.8037.97 for Windows and macOS, along with version 154.0.8037.97 for Linux. One vulnerability, CVE-2026-103628, received a Critical rating due to an out-of-bounds memory write in WebGL that was originally reported in August. Nine additional issues rated High severity affect components such as FileSystem, Compositing, Skia, FedCM, SVG, MediaStream, and WebRTC, including a buffer overflow tracked as CVE-2026-103631. The update also resolves a type confusion flaw in the V8 scripting engine and one Medium-severity issue. Google plans a gradual rollout over the coming days and weeks.
Browser Built on Mistakes: How Real-World Attacks Shaped Modern Browser Defenses
Browser security features such as process isolation, sandboxing, and restrictions on code execution were not designed in isolation but evolved directly in response to concrete attacks over more than a decade. Early threats like malicious Flash advertisements in 2015 demonstrated how a single compromised banner could compromise an entire system, prompting the industry to phase out plugins entirely. Later discoveries, including the Spectre vulnerability, forced browsers to implement stricter site isolation and timing-attack mitigations that remain in place today. Session hijacking and malicious browser extensions further drove the adoption of stronger cookie protections and permission models. BI.ZONE analysts trace this history through specific incidents to show why current architectures prioritize separation of sites into distinct processes. The resulting design reduces the blast radius of any single exploit and continues to adapt as new attack classes emerge.