216.73.217.22

CVE-2026-22807

· Published 21/01/2026 22:15 · Modified 21/01/2026 22:15

Labels: CVE-2026-22807 2026-01-21CVE-2026-22807CWE-94[email protected]

Essential information

Published
21/01/2026 22:15
Modified
21/01/2026 22:15
Author
Creator
CVSS
8.8 HIGH (v3.1)
CISA KEV
No
CWE
CVSS vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

CVSS metrics

Description

vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging Face `auto_map` dynamic modules during model resolution without gating on `trust_remote_code`, allowing attacker-controlled Python code in a model repo/path to execute at server startup. An attacker who can influence the model repo/path (local directory or remote Hugging Face repo) can achieve arbitrary code execution on the vLLM host during model load. This happens before any request handling and does not require API access. Version 0.14.0 fixes the issue.

NVD status

Status
Received — CVE has been recently published to the CVE List and has been received by the NVD.
Source
[email protected]
NVD
View on NVD

Affected products (CPE)

ProductCPE
vllm / vllm cpe:2.3:a:vllm:vllm:0.10.1-0.14.0:*:*:*:*:*:*:*

References