CVE-2026-22807

πŸ”΄ HIGH

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 resolu...

Published
Jan 21, 2026
Last Modified
Jan 26, 2026
Views
57
Bookmarks
0

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.

CVSS Scores

CVSS 3.1 8.8
8.8
HIGH
CVSS 2.0 8.8

Additional Information

Source
security-advisories@github.com
State
Undergoing analysis

Share CVE-2026-22807

Share on Social Media

Copy Link

Embed Code

Request Expert Analysis

Request a professional security analysis for CVE-2026-22807 from our verified experts.

Credits System

Use your credits to get expert analysis from verified security professionals. Purchase more credits anytime!

Add 3 credits for accelerated delivery

Base Cost: 8 credits
Priority Upgrade: + credits
SLA Acceleration: +3 credits
Total Cost:
Your Balance:

Insufficient Credits

You need more credits to submit this request.

Buy Credits

Report Analysis