CVSS 3.1
9.3 criticalCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:NCritical (9.3) in MLflow MLflow. In CISA KEV since 2026-08-19. EPSS 10% estimated probability of exploitation activity in the next 30 days. Fixed in mlflow 3.15.0.
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:Nmlflow (PyPI): fixed in 3.15.0Explain CVE-2026-64849 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-64849?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-64849?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-64849 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.
From the NVD's CPE matches.
| Product | Affected |
|---|---|
| lfprojects:mlflow | from 3.3.0 before 3.15.0 |
| Advisory | Ecosystem | Package | Affected | Fixed in |
|---|---|---|---|---|
| GHSA-7gwp-5pfp-969j | PyPI | mlflow | >=3.3.0 <3.15.0 | 3.15.0 |
| PYSEC-2026-3687 | PyPI | mlflow | <3.15.0 | 3.15.0 |
Red Hat rates this Important.
This flaw is in MLflow Tracking Server webhook delivery. An attacker who can reach the MLflow API can abuse POST /api/2.0/mlflow/webhooks/{id}/test to trigger SSRF and read upstream response bodies. Upstream rates this against the default unauthenticated mlflow server (PR:N).
For OpenShift AI, the operator-managed odh-mlflow-rhel9 image is the primary exposure; the operator enables kubernetes-auth by default, so unauthenticated abuse applies only where MLflow is deployed without authentication. The other 19 RHOAI images embed the mlflow Python package as a client library and do not execute the vulnerable server webhook path in their default role concluding those 19 RHOAI images as Not Affected.
Impact is set to Important with PR:L reflecting typical authenticated RHOAI deployments.
To reduce the attack surface for this vulnerability, restrict network access to the MLflow server. Implement firewall rules or network access controls to limit connectivity to the MLflow instance from untrusted networks. This operational control helps prevent unauthenticated attackers from reaching the vulnerable webhook test endpoint.
| Product | Package | State | Advisory |
|---|---|---|---|
| Red Hat OpenShift AI 3.4 | rhoai/odh-mlflow-rhel9:1787226790 | Fixed 2026-08-27 | RHSA-2026:60520 |
| Red Hat OpenShift AI 3.5 | rhoai/odh-latency-predictor-prediction-rhel9:1786552271 | Fixed 2026-08-26 | RHSA-2026:60367 |
| Red Hat OpenShift AI 3.5 | rhoai/odh-latency-predictor-training-rhel9:1786552250 | Fixed 2026-08-26 | RHSA-2026:60367 |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-datascience-cpu-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-pytorch-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-pytorch-llmcompressor-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-pytorch-rocm-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-th06-cpu-torch210-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-th06-cuda130-torch210-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-th06-rocm64-torch291-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-training-cuda128-torch29-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-codeserver-datascience-cpu-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-datascience-cpu-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-pytorch-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-pytorch-llmcompressor-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-pytorch-rocm-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-tensorflow-cuda-py312-rhel9 | Not affected | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-workbench-jupyter-tensorflow-rocm-py312-rhel9 | Not affected | |
| 1 more in the JSON and on Red Hat's page. | |||
Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-64849.
Repositories on GitHub named after this CVE or listed for it by the PoC-in-GitHub index. For those marked Harbor Check: matches CVE, Flawharbor read the code and found it implements this CVE's exploitation technique: read, not run, and its dependencies only checked for warning signs, so a pointer and no guarantee. The rest are unverified: some hold a working proof of concept, some nothing useful, and some malware made to look like one. Read the code before you run any of it. How Harbor Check reads them, and what it found.
Ways to detect this CVE, not exploit code: the nuclei template that checks a system for it, and the Suricata and Snort rules that name it. A rule set may ship a rule turned off, so check yours by its sid.
The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.