CVE-2026-2614 is a critical security vulnerability in the mlflow component of Red Hat OpenShift AI (versions >=3.4.0 <3.4.2). The vulnerability arises from improper validation in the _create_model_version() handler, where an attacker can include a crafted tag (mlflow.prompt.is_prompt) in a CreateModelVersion request to bypass source path validation. This flaw permits specifying an arbitrary local filesystem path as the model version source. Subsequently, the get_model_version_artifact_handler() serves files from this unverified source, enabling an unauthenticated remote attacker to read arbitrary files on the server, resulting in a complete confidentiality breach. The vulnerability is associated with CWE-22 (Path Traversal) and other weaknesses. Red Hat has released Red Hat OpenShift AI 3.4.2 to fix this issue. No known exploits in the wild have been reported. The vendor advisory recommends upgrading to 3.4.2 and provides documentation for cluster upgrade procedures.

The vulnerability allows an unauthenticated remote attacker to read arbitrary files from the server's filesystem by bypassing source path validation in the mlflow component. This leads to a complete confidentiality compromise of sensitive data stored on the server. There is no impact on integrity or availability reported. The flaw could expose sensitive information that may be used for further attacks or data breaches within Red Hat OpenShift AI environments.

Red Hat has released an official fix in Red Hat OpenShift AI version 3.4.2. Users should upgrade to this version following the vendor's documented upgrade instructions to fully remediate the vulnerability. The vendor advisory states that other mitigation options are either unavailable or do not meet Red Hat's criteria for ease of use, applicability, or stability. Therefore, upgrading is the recommended and effective mitigation.