AI Threat Alert
CVE-2026-33865: MLflow: stored XSS via MLmodel YAML artifact upload
GHSA-fh64-r2vc-xvhr MEDIUM PoC AVAILABLE CISA: TRACK*MLflow's web interface fails to sanitize YAML-based MLmodel artifact files, allowing an authenticated attacker to embed JavaScript payloads that silently execute in any user's browser when they view the artifact in the UI. In organizations where MLflow is shared across data science teams — the norm, not the exception — a compromised low-privilege contributor account is sufficient to hijack the session of an ML engineer or platform administrator who reviews model artifacts as part of routine operations. While no public exploit or CISA KEV listing exists for this CVE, the authenticated-access barrier is low in collaborative MLOps environments where many contributors hold upload rights. Upgrade beyond MLflow 3.10.1 once a patched release ships (track GitHub PR #21435), restrict artifact upload permissions to vetted principals, and enforce a Content Security Policy on all MLflow deployments as an immediate compensating control.
What is the risk?
Medium-High within AI/ML environments. The authentication prerequisite lowers immediate blast radius, but MLflow instances are routinely shared across data science teams with broad contributor access, making the effective barrier to exploitation low in practice. Session hijacking via stored XSS can escalate to full MLflow admin access, enabling model registry manipulation or lateral movement into connected ML infrastructure. No CVSS score or EPSS data is available given the CVE's recency, but CWE-79 in a shared-access MLOps platform with insider-threat relevance warrants prompt remediation.
What systems are affected?
| Package | Ecosystem | Vulnerable Range | Patched |
|---|---|---|---|
| MLflow | pip | <= 3.10.1 | 3.11.1 |
Do you use MLflow? You're affected.
How severe is it?
What should I do?
6 steps-
Patch: Upgrade MLflow beyond 3.10.1 as soon as a fixed release is available — monitor GitHub PR #21435 for merge and release status.
-
Restrict permissions: Limit MLmodel artifact upload rights to trusted, vetted principals; review current contributor lists.
-
Network isolation: Ensure MLflow UI is not exposed to untrusted networks; require VPN or internal-only access.
-
Content Security Policy: Enforce a strict CSP header on the MLflow server to block exfiltration to external domains even if XSS fires.
-
Detection: Monitor web server logs for unexpected outbound requests from the MLflow process; alert on unusual session behavior for admin accounts.
-
Workaround: If immediate patching is not feasible, disable artifact UI access for untrusted users or restrict the feature until a fix is applied.
What does CISA's SSVC say?
Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.
How is it classified?
Which compliance frameworks are affected?
This CVE is relevant to:
Frequently Asked Questions
What is CVE-2026-33865?
MLflow's web interface fails to sanitize YAML-based MLmodel artifact files, allowing an authenticated attacker to embed JavaScript payloads that silently execute in any user's browser when they view the artifact in the UI. In organizations where MLflow is shared across data science teams — the norm, not the exception — a compromised low-privilege contributor account is sufficient to hijack the session of an ML engineer or platform administrator who reviews model artifacts as part of routine operations. While no public exploit or CISA KEV listing exists for this CVE, the authenticated-access barrier is low in collaborative MLOps environments where many contributors hold upload rights. Upgrade beyond MLflow 3.10.1 once a patched release ships (track GitHub PR #21435), restrict artifact upload permissions to vetted principals, and enforce a Content Security Policy on all MLflow deployments as an immediate compensating control.
Is CVE-2026-33865 actively exploited?
Proof-of-concept exploit code is publicly available for CVE-2026-33865, increasing the risk of exploitation.
How to fix CVE-2026-33865?
1. Patch: Upgrade MLflow beyond 3.10.1 as soon as a fixed release is available — monitor GitHub PR #21435 for merge and release status. 2. Restrict permissions: Limit MLmodel artifact upload rights to trusted, vetted principals; review current contributor lists. 3. Network isolation: Ensure MLflow UI is not exposed to untrusted networks; require VPN or internal-only access. 4. Content Security Policy: Enforce a strict CSP header on the MLflow server to block exfiltration to external domains even if XSS fires. 5. Detection: Monitor web server logs for unexpected outbound requests from the MLflow process; alert on unusual session behavior for admin accounts. 6. Workaround: If immediate patching is not feasible, disable artifact UI access for untrusted users or restrict the feature until a fix is applied.
What systems are affected by CVE-2026-33865?
This vulnerability affects the following AI/ML architecture patterns: MLOps platforms, model registry, ML experiment tracking, training pipelines.
What is the CVSS score for CVE-2026-33865?
No CVSS score has been assigned yet.
What is the AI security impact?
Affected AI Architectures
MITRE ATLAS Techniques
AML.T0011.000 Unsafe AI Artifacts AML.T0012 Valid Accounts AML.T0025 Exfiltration via Cyber Means AML.T0049 Exploit Public-Facing Application Compliance Controls Affected
What are the technical details?
Original Advisory
MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated attacker can upload a malicious MLmodel file containing a payload that executes when another user views the artifact in the UI. This allows actions such as session hijacking or performing operations on behalf of the victim. This issue affects MLflow version through 3.10.1
Exploitation Scenario
An attacker holding any MLflow contributor account crafts a malicious MLmodel YAML file embedding a JavaScript payload — for example, a script that exfiltrates the victim's session cookie to an attacker-controlled server via a fetch() call. They register this as a new model version attached to a legitimate-looking experiment. When an MLflow administrator performs a routine artifact review in the web UI, the payload executes silently in their browser. The attacker receives the admin session token, authenticates as admin, and can promote malicious model versions to production, exfiltrate all registered models and experiment data, or alter training run configurations for downstream pipeline compromise.
Weaknesses (CWE)
CWE-79 Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting')
Primary
CWE-79 Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting')
Primary
CWE-79 — Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting'): The product does not neutralize or incorrectly neutralizes user-controllable input before it is placed in output that is used as a web page that is served to other users.
- [Architecture and Design] Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid [REF-1482]. Examples of libraries and frameworks that make it easier to generate properly encoded output include Microsoft's Anti-XSS library, the OWASP ESAPI Encoding module, and Apache Wicket.
- [Implementation, Architecture and Design] Understand the context in which your data will be used and the encoding that will be expected. This is especially important when transmitting data between different components, or when generating outputs that can contain multiple encodings at the same time, such as web pages or multi-part mail messages. Study all expected communication protocols and data representations to determine the required encoding strategies. For any data that will be output to another web page, especially any data that was received from external inputs, use the appropriate encoding on all non-alphanumeric characters. Parts of the same output document may require different encodings, which will vary depending on whether the output is in the: etc. Note that HTML Entity Encoding is only appropriate for the HTML body. Consult the XSS Prevention Cheat Sheet [REF-724] for more details on the types of encoding and escaping that are needed. HTML body Element attributes (such as src="XYZ") URIs JavaScript sections Casca
Source: MITRE CWE corpus.
References
Timeline
Related Vulnerabilities
CVE-2025-15379 10.0 MLflow: RCE via unsanitized model dependency specs
Same package: mlflow CVE-2023-3765 10.0 MLflow: path traversal allows arbitrary file read
Same package: mlflow CVE-2026-2635 9.8 mlflow: security flaw enables exploitation
Same package: mlflow CVE-2023-1177 9.8 MLflow: path traversal allows arbitrary file read/write
Same package: mlflow CVE-2023-2780 9.8 MLflow: path traversal allows arbitrary file read/write
Same package: mlflow