A flaw in Data Science Pipelines allows a restricted tenant to specify a more privileged ServiceAccount when creating a workflow run, bypassing authorization checks. This confused deputy vulnerability lets the tenant run containers with elevated privileges, enabling disclosure of secrets and execution of commands in other users' pods. The issue is classified as CWE-639 and can lead to significant confidentiality and integrity compromise without impacting availability.

The vulnerability affects Red Hat OpenShift AI (RHOAI), specifically the Data Science Pipelines component. No specific version information is listed; affected installations should check the vendor documentation for affected releases.

The CVSS score of 7.1 indicates a high severity, although the EPSS score is not available and the CVE is not listed in CISA’s KEV catalog. Likely attack vectors involve submitting a CreateRun request with a privileged ServiceAccount as a tenant; the attacker only needs tenant-level API access. Once exploited, the attacker gains elevated rights within the cluster and could access sensitive data or compromise other pods. Due to the lack of universally enforced authorization, the risk is elevated for environments that expose workflow creation to multiple tenants.