Skip to main content

Critical Vulnerability in Red Hat OpenShift AI (RHOAI) (CVE-2026-15378)

· 2 min read · SecurityXP

Key Facts

  • CVE-2026-15378 affects the guardrails-detectors component in Red Hat OpenShift AI.
  • A specially crafted XML Schema Definition (XSD) string can trigger a blind server-side request forgery (SSRF).
  • Red Hat rates the flaw CVSS 9.3 (Critical) and ships an official fix.

TL;DR

  • CVE-2026-15378 is a critical blind server-side request forgery (SSRF) flaw in the guardrails-detectors component of Red Hat OpenShift AI.
  • A remote attacker can supply a crafted XML Schema Definition (XSD) string that causes the server to issue requests to attacker-controlled destinations.
  • Red Hat rates the vulnerability CVSS 9.3 (Critical) and provides fixed packages. Organizations running OpenShift AI should update according to the vendor advisory.

What happened

Red Hat published a security fix for CVE-2026-15378 in the guardrails-detectors component used by OpenShift AI. The component processes XSD input during schema validation; a malicious document can abuse that parsing path to make the server perform outbound requests chosen by the attacker.

Because the request is blind, an attacker may not see the response directly, but the server can still be used to reach internal services, cloud metadata endpoints, or other network resources that are not exposed to the internet.

Technical details

The flaw is classified as a server-side request forgery (SSRF). The vulnerable path accepts an XSD document and resolves references while validating it. An attacker who can submit such a document can influence where the server connects.

Red Hat’s advisory lists an official fix and a CVSS v3.1 base score of 9.3. The NVD entry tracks the CVE and links back to the vendor record.

Impact

A successful SSRF can expose internal network topology, bypass network segmentation, and reach services that trust requests originating from the OpenShift AI deployment. The exact blast radius depends on the network position of the affected component and the permissions of its service account.

Remediation

  • Apply the fixed OpenShift AI packages referenced in the Red Hat advisory.
  • Restrict outbound network access from AI workloads to the destinations they actually need.
  • Block access to cloud instance-metadata endpoints from application pods.
  • Review guardrails-detectors inputs and logs for unexpected outbound requests.

Analysis

AI platforms increasingly ingest user-controlled documents and schemas, which makes server-side parsing a high-value attack surface. SSRF flaws in these components can turn a model-serving cluster into a pivot point into the wider environment. Patching is the first step; network egress controls and metadata protections reduce the impact of the next parser bug.

Tags: CVE-2026-15378 Red Hat OpenShift AI SSRF Infrastructure

Sources & References

S SecurityXP
SecurityXP Cybersecurity News & Analysis

SecurityXP delivers cybersecurity news, vulnerability analysis, data breach reports, and threat intelligence.

Security Digest

The most important cybersecurity stories, CVE alerts, and threat intelligence — one concise email. Unsubscribe anytime.

Related Articles