MLflow patches stored XSS and authorization bypass flaws
MLflow versions up to 3.10.1 contain stored XSS and authorization bypass vulnerabilities allowing artifact theft and session hijacking.
Sourcecert.pl — open the original
What to check
- Applies to
- The 2 CVEs named below. Whether you run an affected version is not assessed: check the versions the source lists.
- Exploitation
- No exploitation in the wild and no public exploit is recorded for the CVEs named.
CERT Polska coordinated disclosure of two vulnerabilities in MLflow, an open-source machine learning lifecycle platform. CVE-2026-33865 is a Stored Cross-Site Scripting vulnerability arising from unsafe parsing of YAML-based MLmodel artifacts in the web interface. An authenticated attacker can upload a malicious MLmodel file containing JavaScript payloads that execute when another user views the artifact through the UI, enabling session hijacking or unauthorized actions performed under the victim's context.
CVE-2026-33866 is an authorization bypass vulnerability affecting the AJAX endpoint responsible for downloading saved model artifacts. The endpoint lacks proper access-control validation, allowing users without appropriate permissions to directly query the endpoint and retrieve model artifacts from experiments they should not be able to access. This represents a significant confidentiality breach for organizations using MLflow to manage proprietary machine learning models.
Both vulnerabilities affect all MLflow versions through 3.10.1. Organizations using MLflow should prioritize patching to versions beyond 3.10.1 once available, restrict access to MLflow instances, and review audit logs for unauthorized artifact access or suspicious file uploads.
Mentioned in this report
Source reporting: https://cert.pl/en/posts/2026/04/CVE-2026-33865
Free public brief. Slack alerts on the vendors you run come with the paid plans, and a new account gets three days of them, no card.
What this brief leaves out
This page is the free tier: the write-up, the severity, the names. In the app the same report carries live profiles of the CVEs it names. The paid plans add the security research and news reporting the free tier leaves out.
It sits in a corpus of 16,433 reports from 149 sources, 661 of them written in the last seven days, and it grows through the day.
A new account starts with three days of all of it, no card, then keeps the government and CERT reporting free. Slack alerts are a paid-plan feature: name the vendors you run and Slack hears about the reports that touch them, not the rest of the week's. Nothing is charged unless you subscribe.
Create a free account What it costs