A flaw in the lm-evaluation-harness component of Red Hat OpenShift AI allows an authenticated user who can submit evaluation jobs to provide a malicious model URL in the JobSpec. This design flaw causes the tenant’s model authentication secret to be forwarded to the external URL and also enables the job pod to perform Server‑Side Request Forgery against internal network endpoints.

The affected product is Red Hat OpenShift AI (component lm-evaluation-harness). No specific version information is supplied in the CVE data; organizations should check if their deployment includes this component.

The CVSS score of 6.5 indicates moderate severity, and the EPSS score of less than 1% shows a low probability of exploitation in the wild. The vulnerability is not listed in the CISA KEV catalog. Exploitation requires an authenticated user with job‑submission privileges, meaning it is an insider‑or‑compromised‑user threat. An attacker could obtain sensitive model credentials and make the job pod reach arbitrary internal services, potentially exposing further secrets or providing a foothold for lateral movement.