CVE-2026-22807

EUVD-2026-3678
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.
Code Injection
ProviderTypeBase ScoreAtk. VectorAtk. ComplexityPriv. RequiredVector
NISTPrimary
8.8 HIGH
NETWORK
LOW
NONE
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Base Score
CVSS 3.x
EPSS Score
Percentile: 51.87%
Affected Products (NVD)
VendorProductVersion
vllmvllm
0.10.1 ≤
𝑥
< 0.14.0
𝑥
= Vulnerable software versions
Early Detection
Affected products identified ahead of NVD analysis through intelligence sources.
VendorProductVersionSource
Red HatRed Hat AI Inference Server 3.2
1772160593 ≤
𝑥
< *
ADP
Red HatRed Hat AI Inference Server 3.2
1772160625 ≤
𝑥
< *
ADP
Red HatRed Hat AI Inference Server 3.3
1782352919 ≤
𝑥
< *
ADP
Red HatRed Hat AI Inference Server 3.3
1782353093 ≤
𝑥
< *
ADP
Red HatRed Hat AI Inference Server 3.3
1782352847 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 2.25
1776259063 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 2.25
1772093278 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 2.25
1783998774 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 2.25
1783998857 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 3.3
1770956034 ≤
𝑥
< *
ADP
Red HatRed Hat OpenShift AI 3.4
1772204861 ≤
𝑥
< *
ADP