CVE-2026-6859

HighCVSS 8.8

Published: 22/04/2026(UTC)Updated: 15/07/2026(UTC)Translated: 16/07/2026(UTC)NVD NIST

Exploitation Probability (EPSS)

Low risk

0.42%

34th percentile - higher than 34% of all known CVEs

Summary

The `linux_train.py` script in InstructLab 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.

Risk Assessment

This vulnerability can lead to complete system compromise, potentially resulting in data theft, malware installation, or full control over the organization's IT infrastructure.

Recommendation

Immediately update InstructLab to a version that does not default to `trust_remote_code=True`, or manually set this option to `False` in the `linux_train.py` script.

Other vulnerabilities in InstructLab

CVE-2026-6855High
A flaw was found in InstructLab that allows a local attacker to exploit a path traversal vulnerability in the chat session handler by manipulating the `logs_dir` parameter. This enables the attacker to create new directories and write files to arbitrary locations on the system.

Original NVD description (English source)

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.