CVE-2026-22807
HIGH
GHSA-2pc9-4j83-qjmr
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Lifecycle Timeline
3Tags
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.
Analysis
vLLM is an inference and serving engine for large language models (LLMs). [CVSS 8.8 HIGH]
Sign in for full analysis, threat intelligence, and remediation guidance.
Remediation
Within 7 days: Identify all affected systems running version 0.10.1 and and apply vendor patches promptly. Validate that input sanitization is in place for all user-controlled parameters.
Sign in for detailed remediation steps.
Priority Score
Vendor Status
Share
External POC / Exploit Code
Leaving vuln.today