CVE-2026-6859
The `linux_train.py` script hardcodes `trust_remote_code=True` when loading models from HuggingFace.
Does this matter?
High impact if exploited, but EPSS currently rates exploitation as unlikely (0.44%). Schedule it in the normal patch cycle and watch for a rise in EPSS or a public exploit.
Description
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.
- CVSS 3.1
- 8.8 HIGHCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- EPSS
- 0.44% probability · 37th percentile
- CISA KEV
- Not listed
- Weakness
- CWE-829
- Affected
- redhat/instructlab · redhat/enterprise linux ai
- Source
- secalert@redhat.com
References
- https://access.redhat.com/security/cve/CVE-2026-6859Vendor Advisory
- https://bugzilla.redhat.com/show_bug.cgi?id=2459998Issue Tracking, Vendor Advisory
- https://access.redhat.com/security/cve/CVE-2026-6859Vendor Advisory
- https://bugzilla.redhat.com/show_bug.cgi?id=2459998Issue Tracking, Vendor Advisory
- https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-6859.json
Source: NVD record, EPSS from FIRST.org, KEV from CISA. Refreshed daily.