CVE-2026-6859

A flaw was found in InstructLab. The `linux_train.py` script hardcodes `trust_remote_code=True` when loading models from HuggingFace. This allows a remote attacker to achieve arbitrary Python code execution by convincing a user to run `ilab train/download/generate` with a specially crafted malicious model from the HuggingFace Hub. This vulnerability can lead to complete system compromise.
Configurations

Configuration 1 (hide)

OR cpe:2.3:a:redhat:instructlab:-:*:*:*:*:*:*:*
cpe:2.3:o:redhat:enterprise_linux_ai:3.0:*:*:*:*:*:*:*

History

26 Oct 2026, 21:10

Type Values Removed Values Added
Summary
  • (es) Se encontró un fallo en InstructLab. El script “linux_train.py” codifica de forma rígida “trust_remote_code=True” al cargar modelos de HuggingFace. Esto permite a un atacante remoto lograr la ejecución arbitraria de código Python al convencer a un usuario de ejecutar “ilab train/download/generate” con un modelo malicioso especialmente diseñado del HuggingFace Hub. Esta vulnerabilidad puede llevar a un compromiso completo del sistema.

Information

Published : 2026-04-22 14:17

Updated : 2026-10-26 21:10


NVD link : CVE-2026-6859

Mitre link : CVE-2026-6859

CVE.ORG link : CVE-2026-6859


JSON object : View

Products Affected

redhat

  • enterprise_linux_ai
  • instructlab
CWE
CWE-829

Inclusion of Functionality from Untrusted Control Sphere