A flaw exists in the ServiceAccounts used by the maas-api and maas-controller components of Red Hat OpenShift AI. These accounts are granted permissions that extend across the entire cluster, far beyond what is necessary for their intended functions. If an attacker gains control of one of these ServiceAccounts—by exploiting a remote code execution vulnerability, or by running a malicious pod in the same namespace—he can abuse the excess permissions to create new ClusterRoleBindings that grant cluster‑administrator rights, or to read any secret stored in the cluster. The weakness is a classic example of excessive privilege (CWE‑266).

Red Hat OpenShift AI, specifically the maas-api and maas-controller ServiceAccounts. No specific product versions are listed in the advisory.

The CVSS score of 7.9 indicates high severity. The EPSS score is not available, but because the vulnerability requires compromise of a ServiceAccount that already has broad rights, the exploitation window is significant. The flaw is not listed in the CISA KEV catalog. Exploitation could occur via remote code execution; alternatively, a malicious pod running in the same namespace can use the ServiceAccount’s token to elevate privileges. The attack vector therefore relies on either an initial remote vulnerability or the ability to deploy a pod within the namespace containing the over‑privileged ServiceAccount.