CVE-2026-126994

An EEG model configuration could select code to run during loading

Braindecode project · EEG model configuration loading

First documented disclosure · Review checkpoint

Summary

The loader treated a configuration value as a module and callable to load within its Python process. Code available in the local Python environment could execute under the model loader’s privileges.

Severity and exploitation evidence

Medium · 5.3/10 · CVSS 4.0. Severity source.

Exploitation not established: The reporter demonstrated the behavior on 1.8.1. The reviewed sources do not identify a malicious deployment.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The loader treated a configuration value as a module and callable to load within its Python process.

Conditions required

A user or service loaded an untrusted Braindecode model configuration.

Potential impact

Code available in the local Python environment could execute under the model loader’s privileges.

Affected versions and configurations

The researcher verified 1.8.1 and found the code in tags 1.4.0 through 1.8.1. The CNA’s structured range lists 1.8.0 and 1.8.1.

Fix and mitigation status

2.0.0.

Practical next steps

Upgrade Braindecode to 2.0.0 or later and load untrusted model artifacts only under restricted privileges.

Operating lesson

Treat model configuration that can select executable objects like code.

Evidence and limits

The CVE record was published on 2026-11-12.

Sources