CVE-2026-4944: Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control

Severity: high (CVSS 8.8 v3.0)
Exploited in the wild: not listed in CISA KEV
EPSS: 0.7% probability of exploitation in the next 30 days
Public exploit or detection template: none known
Published: 2026-05-28 · updated 2026-06-17
Weakness: CWE-22

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.

Affected products

vllm-project vllm: through latest

References

Get alerted when CVE-2026-4944 or the next vulnerability affects your stack