Executive brief
MLflow is a platform used by data scientists to manage machine learning lifecycles, including the execution of automated tasks or 'jobs.' A security flaw allows unauthorized users to bypass login requirements and interact with these job management features. This could allow an attacker to run malicious code on the server, steal sensitive data from job results, or disrupt operations by deleting or spamming tasks.
Technical details
A vulnerability exists in MLflow's FastAPI implementation where endpoints under `/ajax-api/3.0/jobs/*` fail to enforce authentication or authorization checks even when `basic-auth` is enabled. If the server is configured with `MLFLOW_SERVER_ENABLE_JOB_EXECUTION=true` and has allowlisted job functions, an unauthenticated remote attacker can submit, read, search, or cancel jobs. This bypass can lead to Remote Code Execution (RCE) if the allowlisted jobs perform privileged operations like shell execution. It also facilitates Denial of Service (DoS) through job spamming and unauthorized data exposure.
Affected products
- MLflow MLflow latest
- Red Hat Red Hat OpenShift AI (RHOAI)
Timeline
- 2026-04-03: disclosed
- 2026-04-03: advisory