CVE-2026-13717 is a vulnerability in the Red Hat OpenShift AI (RHOAI) MaaS Gateway caused by improper access control due to insecure default configuration. In a model-serving context, this flaw allows a standard user with low privileges to intercept, read, log, and alter all MaaS model traffic, including sensitive information like access keys, input prompts, and outputs. The vulnerability affects any namespace authorized to use the MaaS Gateway, enabling namespace users to hijack shared model-serving traffic. The root cause is an overly permissive default setting for allowedRoutes.namespaces.from, which allows all namespaces access. This vulnerability is tracked as CWE-284 (Improper Access Control) and has a CVSS v3.1 base score of 8.8 (high severity) with network attack vector, low complexity, low privileges required, no user interaction, and high impact on confidentiality, integrity, and availability.

The vulnerability enables a low-privileged user within an authorized namespace to intercept and manipulate all traffic passing through the MaaS Gateway, including sensitive data such as access keys, input prompts, and model outputs. This results in significant information disclosure and data tampering, potentially compromising the confidentiality, integrity, and availability of the model-serving infrastructure and its data.

A fix is available in Red Hat OpenShift AI version 3.4.3. Users should upgrade to this version following Red Hat's official documentation to fully apply the errata update. As a mitigation, Gateway access can be restricted to specific namespaces, but this does not prevent users within authorized namespaces from hijacking traffic of others. Fully locking down access to the MaaS Gateway is possible but may defeat the component's self-service design. Therefore, upgrading to the fixed version is the recommended remediation.