Tools AI Risk Radar ai-incident-0259
Incident record
MLflow fixes unauthenticated requests to internal services
- Severity
- Medium
- Status
- Patched
- Type
- Infra Vuln
- Target
- MLflow tracking server with optional TruLens integrations
- Actor
- Researcher disclosure
- CVE
- CVE-2026-115886
What happened
A caller could make an affected MLflow server send requests to an address they chose, including an internal service. The vulnerable setup requires both optional trulens and trulens-providers-litellm packages.
The maintainer’s advisory was published on 22 October; the CVE record followed on 23 October. Affected versions run from 3.12.0rc0 up to, and excluding, 3.17.0. Version 3.17.0 fixes the issue. The vulnerable scorer endpoint accepted a caller-selected class and constructor arguments before its normal tracking-store work. The demonstrated request carried the attacker’s own chosen headers.
Sources
- MLflow maintainer advisorygithub.com/mlflow/mlflow/security/advisories/GHSA-2wqp-q3c6-…
- MLflow 3.17.0 releasegithub.com/mlflow/mlflow/releases/tag/v3.17.0
One record from the AI Risk Radar, maintained by the Security Desk.Data: CSV · JSON ·RSS · CC BY 4.0 with attribution to YFarmX.Tracker updated · 28 October 2026