MLflow - RCE (CVE-2026-116692)
Description
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, third-party scorer deserialization in mlflow/genai/scorers/base.py accepts request-selected descendants of THIRD_PARTY_SCORER_ALLOWED_MODULES rather than requiring an exact allowed module. A caller can use MLflow artifact endpoints to place a Python module beneath a writable allowed package and submit its module and class through /ajax-api/3.0/mlflow/scorer/invoke. Scorer.model_validate_json() imports the module and constructs the class before trace processing. Authentication-disabled deployments permit this code execution without credentials, while basic-auth deployments require a principal capable of creating an experiment and run and uploading artifacts. In either configuration, the allowed scorer package must be writable by the server process. This issue is fixed in version 3.17.0.
psychologyCrowdSec Analysis
CVE-2026-116692 is a critical remote code execution vulnerability in MLflow versions 3.12.0rc0 through 3.16.x, caused by unsafe deserialization of third-party scorers. An attacker can place a Python module in a writable allowed package and invoke it through MLflow’s scorer endpoint; authentication-disabled deployments may be exploitable without credentials, while basic-auth deployments require permissions to create experiments and runs and upload artifacts. Successful exploitation can let attackers execute code on the MLflow server, potentially compromising its confidentiality, integrity, and availability. The issue is fixed in MLflow 3.17.0.
CrowdSec has been tracking this vulnerability and its exploits since 18th of November 2026.
CrowdSec has not observed any significant exploitation activity targeting CVE-2026-116692 across its network. As a result, no community-driven trend analysis is available for this vulnerability at this time.
Exploitation targets MLflow endpoints including /api/2.0/mlflow/experiments/create, /api/2.0/mlflow/runs/create, /ajax-api/2.0/mlflow/upload-artifact, and /ajax-api/3.0/mlflow/scorer/invoke. Attackers upload a crafted Python scorer and reference it through the serialized_scorer field to trigger code execution during scorer deserialization.
Full Intelligence Available
Connect your API key to access CrowdSec scores, exploit timelines, and attacker IPs.
| Event | Date |
|---|---|
| CVE Published | 2026-10-27 |
| Rule Released | 2026-11-18 |
Related Weaknesses
The product constructs all or part of a code segment using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the syntax or behavior of the intended code segment.
The product uses external input with reflection to select which classes or code to use, but it does not sufficiently prevent the input from selecting improper classes or code.
Remediation & Protection
External References
- https://github.com/mlflow/mlflow/pull/25906
- https://github.com/mlflow/mlflow/commit/a732443e695c3c677680c54320efc394c81654ab
- https://github.com/mlflow/mlflow/releases/tag/v3.17.0
- https://github.com/mlflow/mlflow/security/advisories/GHSA-26p8-2jq9-3vq9
- https://github.com/projectdiscovery/nuclei-templates/blob/main/http/cves/2026/CVE-2026-116692.yaml