CVE-2026-27893
HIGHvLLM's hardcoded trust_remote_code=True in NemotronVL and KimiK25 bypasses user security opt-out
Title source: cnaDescription
vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the user's explicit `--trust-remote-code=False` security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.
References (3)
Core 3
Core References
X_Refsource_Confirm x_refsource_confirm
https://github.com/vllm-project/vllm/security/advisories/GHSA-7972-pg2x-xr59
X_Refsource_Misc x_refsource_misc
https://github.com/vllm-project/vllm/pull/36192
X_Refsource_Misc x_refsource_misc
https://github.com/vllm-project/vllm/commit/00bd08edeee5dd4d4c13277c0114a464011acf72
Scores
CVSS v3
8.8
EPSS
0.0075
EPSS Percentile
49.9%
Attack Vector
NETWORK
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
CISA SSVC
Vulnrichment
Exploitation
none
Automatable
no
Technical Impact
total
Details
CWE
CWE-693
Status
published
Products (3)
pypi/vllm
0.10.1 - 0.18.0PyPI
vllm/vllm
0.10.1 - 0.18.0
vllm-project/vllm
>= 0.10.1, < 0.18.0
Published
Mar 27, 2026
Tracked Since
Mar 27, 2026