mlflow: remote code execution

mlflow

MLflow 3.12.0 introduced registration and deserialization support for third-party scorers in OSS MLflow. During scorer deserialization, the server accepts a request-controlled Python module path beneath one of four allowed package prefixes, imports that module, obtains a request-controlled class, and instantiates it synchronously.

Do this now

Upgrade mlflow to 3.17.0 and limit who can reach the MLflow server until then.

Fix it

pip install --upgrade "mlflow>=3.17.0"

First seen: 27 Oct 2026
Last update: 27 Oct 2026, 22:17 UTC
Fixed in: mlflow 3.17.0
Identifiers: CVE-2026-116692
CVSS: 9.8
Source: GitHub advisory
On the menu as: ML infra
PyPI packages

Check this repository for exposure to one security incident from pwned.coffee. Report first, change nothing until I say so.

## mlflow: remote code execution
Severity: critical. Patch status: patched.
Affected: mlflow
Fixed in: mlflow 3.17.0
Ids: CVE-2026-116692

MLflow 3.12.0 introduced registration and deserialization support for third-party scorers in OSS MLflow. During scorer deserialization, the server accepts a request-controlled Python module path beneath one of four allowed package prefixes, imports that module, obtains a request-controlled class, and instantiates it synchronously.

## Steps
1. Search every lockfile and manifest (package-lock.json, yarn.lock, pnpm-lock.yaml, requirements*.txt, poetry.lock, uv.lock, composer.lock, go.sum, Cargo.lock, Gemfile.lock) for `mlflow`. For each hit, compare the installed version with the fix on the same major line: mlflow 3.17.0. If it's lower, give the exact upgrade command for the package manager this repo uses.

## Fix by release line
| Line | State | Fix | Until then |
|---|---|---|---|
| mlflow >= 3.12.0rc0 < 3.17.0 | fixed | 3.17.0 | — |

## Report
A table: file, product or package, version in use, affected yes/no, exact fix. Then anything a human has to check outside the repo.

If you have the pwned.coffee MCP server (https://pwned.coffee/mcp), call lookup_cve("CVE-2026-116692") for the latest, and check_repo with this repo's lockfiles, Dockerfiles and workflows for a full receipt.
Sources: https://github.com/advisories/GHSA-26p8-2jq9-3vq9 https://github.com/mlflow/mlflow/commit/a732443e695c3c677680c54320efc394c81654ab https://github.com/mlflow/mlflow/pull/25906 https://github.com/mlflow/mlflow/releases/tag/v3.17.0 https://github.com/mlflow/mlflow/security/advisories/GHSA-26p8-2jq9-3vq9
https://pwned.coffee/i/mlflow-vulnerable-to-remote-code-execution-on-unauthenti-cve-2026-116692

Receipt

Exploitation: No known exploitation
Fix by release line: Product Affected Fixed in Source
mlflow (pip) >= 3.12.0rc0, < 3.17.0 3.17.0 GitHub advisory checked 27 Oct
Score: CVSS 9.8 (critical). CVSS 3.1 9.8 (CNA GitHub_M) for CVE-2026-116692
Last checked: 27 Oct 2026, 22:18 UTC

Product	Affected	Fixed in	Source
mlflow (pip)	>= 3.12.0rc0, < 3.17.0	3.17.0	GitHub advisory checked 27 Oct

Sources: KEV – CVE record ✓ Vendor –

Sources

Wrong fix version, or a source we missed? Send it back to chef@pwned.coffee with the source, and we'll correct it here.

Totalone very hot coffee