CVE-2026-15044 describes a vulnerability in the TrustyAI Service Operator component of Red Hat OpenShift AI (RHOAI). When deploying services such as gorch or NemoGuardrails, failure to enable a particular security setting results in exposed communication channels that do not require user authentication. This exposure allows any program within the Kubernetes cluster to interact with the AI guardrails and orchestrator without proper authorization. Exploitation could lead to unauthorized access to sensitive data and limited changes to AI models. The CVSS 3.1 vector indicates the attack requires adjacent network access with low attack complexity and low privileges, no user interaction, and impacts confidentiality highly, integrity to a lesser extent, and no availability impact. No vendor advisory details on remediation or patches are currently provided.

The vulnerability allows unauthorized intra-cluster programs to access AI guardrails and orchestrator communication channels without authentication. This can lead to unauthorized disclosure of sensitive information and limited unauthorized modifications to AI models. The impact on confidentiality is high, integrity impact is low, and availability is not affected. The attack requires low privileges and adjacent network access within the cluster environment.

Patch status is not yet confirmed — check the Red Hat advisory at https://access.redhat.com/security/cve/CVE-2026-15044 for current remediation guidance. Until an official fix or mitigation is provided, ensure that the specific security setting referenced in the advisory is enabled when deploying services like gorch or NemoGuardrails to prevent exposure of communication channels without authentication.