CVE-2026-116692: MLflow: Remote code execution on unauthenticated MLflow servers via third-party scorer deserialization importing caller-placed artifact modules
- Is it exploited?
Reported, not confirmed by CISA. VulnCheck KEV lists it as exploited in the wild. CISA has not added it to its Known Exploited Vulnerabilities catalog as of .
VulnCheck KEV lists it since . It cites 1 public report, the first from .
- How likely is exploitation in the next 30 days?
- 0.7%
EPSS score, percentile 53.8 of all scored records. A probability estimated by FIRST, not a report of exploitation.
CISA's SSVC assessment in the CVE record lists exploitation as poc.
- How severe?
- 9.8 Critical
CVSS v3.1 base score, given by the CNA
- What should I do?
CISA has set no required action, because the vulnerability is not in KEV.
No patch or vendor advisory is listed in the CVE record. Check MLflow's advisories.
- Which versions are affected?
Affected
- >= 3.12.0rc0, < 3.17.0
Source: CISA KEV 2026.11.10, as of ; VulnCheck KEV, as of ; FIRST EPSS, as of ; CVE List, as of . About the data.
Description
MLflow: Remote code execution on unauthenticated MLflow servers via third-party scorer deserialization importing caller-placed artifact modules
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.
Source: CVE List, as of . About the data.
What to do
Advisories and patches
No patch, vendor advisory or mitigation is listed in the CVE record. Check MLflow's own security advisories for a fix, and compare your version with the affected versions below.
This summary is based on the CVE record as of . Check the vendor's advisory before you act: affected versions and fixes can change.
Affected products and versions
Versions by product, as listed in the CVE record (1 product)
| Status | Versions |
|---|---|
| Affected | >= 3.12.0rc0, < 3.17.0 |
Source: CVE List, as of . About the data.
Weakness
CWE-94, CWE-470
- CWE-94: Improper Control of Generation of Code ('Code Injection')
- CWE-470: Use of Externally-Controlled Input to Select Classes or Code ('Unsafe Reflection')
CWE and CAPEC are maintained by MITRE; the entries are those named in the CVE record.
Timeline
| Date | Event |
|---|---|
| CVE record published | |
| CVE record last updated |
References
4
Known exploited MLflow vulnerabilities
- CVE-2026-64849: MLflow Server-Side Request Forgery Vulnerability