CVE
CVE-2026-4944
- First Reported
- Sep 29, 2026
- Latest Reported
- Sep 29, 2026
CVE Enrichment
- CVSS
- 8.8 HIGH
- EPSS
- 1.00%
- EPSS Percentile
- 61.61%
- CISA KEV
- Not listed
Description
vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.
Technical Details
- CVSS Version
- 3.0
- CNA / NVD Source
- security@huntr.dev
- Published
- 29-May-2026 02:16
- Last Modified
- 17-Jun-2026 17:57
- EPSS Date
- 2026-10-10
- CWE
- CWE-22
CVSS Vector
CVSS:3.0 AV:N /
AC:L /
PR:N /
UI:R /
S:U /
C:H /
I:H /
A:H
References
Sources: NVD · FIRST EPSS · CISA KEV
Latest successful provider update; providers refresh independently.
Reported Context (1)
- at the center of a vulnerability like this, even this year (see LMDeploy bug CVE-2026-46432, vLLM bug CVE-2026-4944, and InstructLab bug CVE-2026-6859). Unsloth Studio Flaw Allowed Malicious Models to Execute Code During Inspection
CVE (2)
People (1)
MITRE ATT&CK (1)
Vendors (2)
Products (6)
Note: Related entities, including threat actors, malware, CVEs, MITRE ATT&CK techniques, vendors, products, tools, countries, and industries, are shown when they appear in the same reporting. Their presence does not necessarily mean they were targeted, compromised, vulnerable, responsible for the activity, or directly involved in the incident.