The vulnerability enables a user who lacks permission to an experiment to retrieve model artifacts through an unsecured AJAX endpoint. Because the application omits required authorization checks, attackers can access confidential data that should be protected, aligning with CWE‑862.

MLflow releases up through version 3.10.1 are affected. All builds of the MLflow product distributed by the Mlflow vendor fall into this range.

The CVSS score of 5.3 indicates a moderate severity, and the issue is not listed in the CISA KEV catalogue. No EPSS score is reported. An attacker can exploit the flaw by issuing an HTTP request to the AJAX endpoint while authenticated with any user account that does not have experiment‑level permissions. The exploit requires no special privileges beyond network access to the MLflow server, making it likely to be performed from an internal network or any system that can reach the endpoint.