CVE-2026-96804 CVE-2026-96804
Remote attackers can execute arbitrary code via crafted MLmodel artifacts due to missing pickle deserialization security controls in MLflow's statsmodel flavor.
- Exploited in the wild
- Not known
- Login needed
- Any signed-in user
- Attackable
- Over the internet
- Victim action needed
- Not needed
- Fix
- None listed
What is CVE-2026-96804?
CVE-2026-96804 is a high-severity unsafe deserialization vulnerability in MLflow, published on 23 Sep 2026. Any signed-in user can exploit it over the network, with no action from the victim.
Official description
MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model(), which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact.
Weakness: CWE-502
Which versions are affected by CVE-2026-96804?
How do I fix CVE-2026-96804?
No fix has been published for CVE-2026-96804 yet. Limit exposure of the affected software until one is available.
Upgrade, and only deserialise data from trusted sources or with an allow-list of types. More about Unsafe deserialization.
Is CVE-2026-96804 being exploited?
There is no known exploitation of CVE-2026-96804 in the wild and no public exploit code.
Sources
Also known as GHSA-2qmw-5xhm-qq5w.