8.8
High
CVE-2026-18982 A flaw was found in the RHOAI training-operator
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.
Affected products and versions
| Product | Affected versions | Fixed in |
|---|---|---|
| Red Hat OpenShift AI (RHOAI) | Versions not published | |
| Red Hat OpenShift AI 3.3 1785188461 | earliest → < 1785188461 | 1785188461 |
| Red Hat OpenShift AI 2.25 1785187053 | earliest → < 1785187053 | 1785187053 |
| Red Hat OpenShift AI 3.4 1787361677 | earliest → < 1787361677 | 1787361677 |
| Red Hat OpenShift AI 3.5 1786552250 | earliest → < 1786552250 | 1786552250 |
Other CVEs of Red Hat OpenShift AI (RHOAI)
The 406 CVEs of Red Hat OpenShift AI (RHOAI) →
CVE-2026-83495
Trainer: trainer: trainjob env passthrough enables valuefrom.secretkeyref injection for namespace secret exfiltration
Not rated
CVE-2026-83494
Training-operator: pytorchjob validateupdate casts oldjob from newobj, bypassing managedby immutability (asvs v5.1.3; upstream bug)
Not rated
CVE-2026-83493
Distributed-workloads: github actions pinned by mutable tag/branch, including third-party snyk/actions/setup@master
Not rated
CVE-2026-107912
Io.quarkus/quarkus-vertx-http: quarkus-vertx-http: authorization bypass in management interface via path concatenation flaw
7.7
High