HIGH8.8

PYSEC-2026-2522

InstructLab Includes Functionality from Untrusted Control Sphere

Details

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.

Are you affected?

Enter the version of the package you're using.

Affected packages

PyPI/instructlab

Introduced in: 0

No fixed version published yet for instructlab (pip). Pin to a known-safe version or switch to an alternative.

References

https://nvd.nist.gov/vuln/detail/CVE-2026-6859[ADVISORY]
https://access.redhat.com/security/cve/CVE-2026-6859[WEB]
https://bugzilla.redhat.com/show_bug.cgi?id=2459998[WEB]
https://github.com/instructlab/instructlab[PACKAGE]
https://pypi.org/project/instructlab[PACKAGE]
https://github.com/advisories/GHSA-rxpq-xgqx-fr7p[ADVISORY]