A high-severity privilege assignment flaw in the Red Hat OpenShift AI 3.3 LMEvalJob controller could allow authenticated attackers to gain unauthorized elevated access.

The vulnerability is an incorrect privilege assignment (CWE-266) within the LMEvalJob controller. The CVSS vector (PR:L) indicates that a low-privileged authenticated user can exploit this flaw to impact confidentiality and integrity.

Successful exploitation allows an authenticated user to perform unauthorized actions or access sensitive data, potentially leading to a compromise of the AI service infrastructure. With a CVSS score of 8.1, this represents a significant risk to organizational data security and system integrity, necessitating prompt remediation to prevent internal abuse.

Immediate Action: Update to Red Hat OpenShift AI 3.3, specifically utilizing the build/release 1785187521 or later, as documented in RHSA-2026:53263.

Proactive Monitoring: Review audit logs for unusual LMEvalJob controller activity or unauthorized privilege escalation patterns within the OpenShift cluster.

Compensating Controls: Implement strict Role-Based Access Control (RBAC) policies within OpenShift to limit the permissions of users capable of interacting with the LMEvalJob controller.

Public Exploit Available: No

The 8.1 CVSS score underscores the high risk posed by this privilege assignment flaw. Administrators should prioritize the application of the vendor-provided security updates to ensure the integrity of the OpenShift AI environment.