CVE-2026-96775

A flaw was found in MLflow's dspy flavor. The security control intended to prevent pickle deserialization is only enforced when the model file path ends with '.pkl'. This oversight allows a remote attacker to bypass the security control and execute arbitrary code by providing a specially crafted...

CVE-2026-96804

A flaw was found in MLflow. The statsmodel component of MLflow fails to enforce a security control that prevents insecure deserialization during model loading. This allows a remote attacker to execute arbitrary code by submitting a specially crafted MLmodel artifact. This vulnerability can lead t...

CVE-2026-96775

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96804

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96804

The MLflow machine learning platform (versions 2.1.0 to 3.14.0 ) contains a vulnerability in its statsmodels flavor component. The _load_model() function fails to enforce the MLFLOW_ALLOW_PICKLE_DESERIALIZATION security control, allowing an attacker to execute arbitrary code by submitting a speci...

CVE-2026-96804

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96804 CVE-2026-96804

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96804 CVE-2026-96804

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

EUVD-2026-85421

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96775

The MLflow machine learning platform is vulnerable to arbitrary remote code execution (RCE) due to a flaw in the dspy flavor (versions >= 2.0 ). The issue stems from a security control bypass where the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False setting is only enforced if the model_path ends in...

EUVD-2026-85410

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96775 MLflow dspy bypasses pickle deserialization control

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96775 MLflow dspy bypasses pickle deserialization control

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96775

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

PT-2026-97413

Name of the Vulnerable Software and Affected Versions MLflow versions 2.0 and later Description The dspy flavor in MLflow incorrectly implements the MLFLOW ALLOW PICKLE DESERIALIZATION=False security control. The control is only applied when the model path ends with the .pkl extension, enabling a...

CVE-2026-96804: Deserialization of Untrusted Data

MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOWALLOWPICKLEDESERIALIZATION=False security control entirely in loadmodel, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...

CVE-2026-96775: Deserialization of Untrusted Data

MLflow's dspy flavor, versions = 2.0, applies the MLFLOWALLOWPICKLEDESERIALIZATION=False security control only when the modelpath ends in .pkl, which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact...