Code Injection in MLflow - CVE-2026-116692
Published: November 9, 2026
Vulnerability details
The vulnerability allows an unauthenticated remote attacker to execute arbitrary Python code with the MLflow server process's privileges on servers with authentication disabled.
The vulnerability exists due to third-party scorer deserialization trusting package-name prefixes when importing caller-selected modules beneath writable scorer packages. An unauthenticated remote attacker can create an experiment and run, upload a malicious Python artifact beneath an allowed package, and invoke the scorer to execute it without user interaction.