Excessive Permissions in Red Hat Openshift AI Service TrustyAI

CVE-2025-12103

What is CVE-2025-12103?

A security flaw in the TrustyAI component of Red Hat Openshift AI Service allows all service accounts and users within a cluster to gain unrestricted access to pod information across all namespaces. This vulnerability arises from the creation of a role, trustyai-service-operator-lmeval-user-role, and a ClusterRoleBinding, trustyai-service-operator-default-lmeval-user-rolebinding, that is inadvertently assigned to system:authenticated. Consequently, every user or service account can list pods in any namespace, as well as access all persistentvolumeclaims and lmevaljobs, creating potential security risks across the platform.

References