Red Hat OpenShift AI Vulnerability Lets Attackers Seize Infrastructure Control

Red Hat OpenShift AI vulnerability, identified as CVE-2025-10725, allows attackers with minimal access to escalate privileges and seize control of entire clusters. The flaw arises from an overly permissive ClusterRoleBinding that links the built-in system:authenticated group to the kueue-batch-user-role, enabling any authenticated user to perform broad job-creation rights across the cluster.

A low-privileged user, such as a data scientist with a standard Jupyter notebook account, can exploit this weakness to gain full cluster administrator rights, according to Red Hat. Once elevated, an attacker can steal sensitive data, disrupt services, and take control of the underlying infrastructure, leading to a complete breach of both the platform and hosted applications.

The advisory recommends immediate mitigation steps, including revoking the ClusterRoleBinding associating kueue-batch-user-role with system:authenticated and assigning job-creation privileges only to specific users or groups, while reviewing other ClusterRoleBindings to prevent other overly broad assignments. Red Hat notes CVE-2025-10725 is formally documented on the CVE website and the NVD, and stresses the importance of least-privilege practices in Kubernetes environments.