MLflow impacted the Experiment‑scoped Label Schema CRUD API, where a missing authorization check allows an attacker to manipulate label schema definitions remotely. The vulnerability is triggered by calling an undocumented function, and exploitation requires a high level of complexity, making it difficult to execute. Nonetheless, the attack can be launched from outside the host, and the vulnerability has been disclosed publicly.

The affected product is MLflow, a machine‑learning platform. The vulnerability exists in all releases up to commit 4666cffc7912ea606d592fc38d6a75e2935f65e7; no specific product versions are listed in the CNA data.

The CVSS score is 2.3, reflecting low overall severity. EPSS is not available, and the vulnerability is not included in CISA’s KEV catalog. The remote attack vector and high complexity suggest that exploitation is unlikely but not impossible; potential impact is limited to unauthorized creation, modification, or deletion of label schemas used in experiments.