CVE-2026-18982

Odh-training-operator-rhel9: rhoai fork aggregates training job create onto native edit/admin clusterroles

Published Aug 10, 2026 CVSS 3.1 · 8.8 High EPSS < 1% KEV · Not listed
Description
A flaw was found in the RHOAI training-operator. This vulnerability allows a user with standard edit or admin roles in any Kubernetes namespace to escalate their privileges. Through the creation of training jobs, an attacker can impersonate service accounts, access the host filesystem, and potentially execute arbitrary code remotely. This issue arises from the aggregation of training job permissions onto native Kubernetes edit and admin ClusterRoles, coupled with unrestricted PodTemplateSpec passthrough.

Analysis

AI-generated

Technical impact

The vulnerability resides in the OpenShift AI training‑operator. By merging the permissions of native Kubernetes edit and admin ClusterRoles into the creation of training jobs and accepting arbitrary PodTemplateSpec objects without validation, any user who holds edit or admin rights in a Kubernetes namespace can create a training job that impersonates a service account. This allows the attacker to access the host filesystem and potentially launch arbitrary code on the host node. The flaw exemplifies a failure of authorization (CWE‑250).

Affected Systems

Affected vendors and products include Red Hat OpenShift AI (RHOAI). Any installation that includes the training‑operator, regardless of the cluster or node version, is vulnerable as the description provides no specific version constraints. Consequently, all current copies of the operator deployed in a cluster are potentially affected.

Risk and Exploitability

The CVSS score of 8.8 indicates high severity. EPSS data is unavailable, but the lack of a KEV listing suggests exploitation is not currently widespread. The attack vector is internal to a Kubernetes cluster: a user with edit or admin privileges in any namespace can trigger the flaw without external network reach, and successful exploitation leads to privilege escalation and remote code execution on the host, giving an attacker full control of the node. The vulnerability is therefore a high‑risk assault vector for privileged users.

Remediation

Vendor Workaround

Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability.


OpenCVE Recommended Actions

  • Apply the vendor‑issued patch for the RHOAI training‑operator as soon as it becomes available.
  • If no patch is yet released, remove or disable the training‑operator by deleting its deployment or disabling its Custom Resource Definition in the cluster.
  • Restrict RBAC by ensuring only a minimal set of users have edit or admin rights in namespaces where the training‑operator runs, or use a dedicated restricted ServiceAccount that isolates host access.

Generated by OpenCVE AI on August 10, 2026 at 23:19 UTC.

Tracking

Sign in to view the affected projects.

Advisories

No advisories available
History
Type Values Removed Values Added
CPEs cpe:/a:redhat:openshift_ai:3.5::el9
References
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'none', 'Technical Impact': 'total'}, 'version': '2.0.3'}

Type Values Removed Values Added
CPEs cpe:/a:redhat:openshift_ai:2.25::el9
cpe:/a:redhat:openshift_ai:3.4::el9
References
Type Values Removed Values Added
CPEs cpe:/a:redhat:openshift_ai cpe:/a:redhat:openshift_ai:3.3::el9
References
Type Values Removed Values Added
First Time appeared Red Hat
Red Hat red Hat Openshift Ai (rhoai)
Vendors & Products Red Hat
Red Hat red Hat Openshift Ai (rhoai)
Type Values Removed Values Added
Description A flaw was found in the RHOAI training-operator. This vulnerability allows a user with standard edit or admin roles in any Kubernetes namespace to escalate their privileges. Through the creation of training jobs, an attacker can impersonate service accounts, access the host filesystem, and potentially execute arbitrary code remotely. This issue arises from the aggregation of training job permissions onto native Kubernetes edit and admin ClusterRoles, coupled with unrestricted PodTemplateSpec passthrough.
Title Odh-training-operator-rhel9: rhoai fork aggregates training job create onto native edit/admin clusterroles
First Time appeared Redhat
Redhat openshift Ai
Weaknesses CWE-250
CPEs cpe:/a:redhat:openshift_ai
Vendors & Products Redhat
Redhat openshift Ai
References
Metrics cvssV3_1

{'score': 8.8, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H'}

Weaknesses
  • CWE-250

    Execution with Unnecessary Privileges

Sources

Status
PUBLISHED
Assigner
redhat
Published
Updated
Sep 21, 2026, 10:25 AM
Reserved
Aug 5, 2026, 4:53 PM
Status
Awaiting Analysis
Published
Aug 10, 2026, 9:17 PM
Modified
Sep 21, 2026, 11:17 AM
Severity
Critical
Public date
Aug 10, 2026, 6:54 PM