CVE-2026-3198 describes an improper access control vulnerability (CWE-284) in MLflow version 3.9.0 when basic-auth is enabled. The issue affects multiple Gateway API list endpoints that fail to enforce proper authorization checks, allowing any authenticated user to access sensitive information such as gateway secrets, API keys, endpoints, and model definitions. The vulnerability does not impact integrity or availability but exposes confidential data to unauthorized users. No official patch or remediation level has been published, and no known exploits are reported in the wild.

The vulnerability allows unauthorized disclosure of sensitive information including API keys and proprietary model details to any authenticated user. This compromises confidentiality but does not affect data integrity or system availability. Exposure of API keys could lead to further unauthorized access if those keys are reused or provide elevated privileges elsewhere.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, restrict access to MLflow instances with basic-auth enabled to trusted users only and monitor for unauthorized access attempts. Avoid exposing MLflow Gateway API endpoints publicly and consider additional access controls or network segmentation to limit exposure.