high8.8
CVE-2026-6859
InstructLab Includes Functionality from Untrusted Control Sphere
Impact
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.
Affected packages
| Ecosystem | Package | Affected | Fixed in |
|---|---|---|---|
| pip | instructlab | <= 0.26.1 | none recorded |
Affected products
| Vendor | Product | Affected versions |
|---|---|---|
| Redhat | Instructlab | All versions |
| Redhat | Enterprise Linux Ai | Version 3.0 |
Weakness classification
References
- NVD record
- CVE.org record
- GitHub advisory
- https://github.com/advisories/GHSA-rxpq-xgqx-fr7p
- https://nvd.nist.gov/vuln/detail/CVE-2026-6859
- https://access.redhat.com/security/cve/CVE-2026-6859
- https://bugzilla.redhat.com/show_bug.cgi?id=2459998
- https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-6859.json