vllm-project · vLLM · published 2026-01-21
CVE-2026-22807vLLM affected by RCE via auto_map dynamic module loading during model initialization
Routine No exploitation evidence; EPSS 1.0%.
CVE-2026-22807 is not known to be exploited in the wild. EPSS estimates a 1.0% chance of exploitation in the next 30 days, higher than 63% of all CVEs. This vLLM vulnerability: CVSS 8.8 (v3.1), a successful attack gives total control of the component. Fixed upstream in vLLM 0.14.0. Routine: patch when convenient.
Jump to: Remediation Exploit intelAffected versionsDescription
Remediation#
Fixed upstream in vLLM 0.14.0.
Patch events
AI toolingvLLM
vLLM security release: 0.14.0
Upgrade to
0.14.0Why: Vendor security release for server software (most severe: CVE-2026-22807).
Exploit intelligence#
| Signal | Value | Source |
|---|---|---|
| Known exploited | No | CISA KEV |
| Ransomware use | — | CISA KEV |
| EPSS (30-day exploitation probability) | 1.0% (63th pct) | FIRST EPSS v2026.06.15 |
| CVSS base score | 8.8 (v3.1)CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H | CNA / CISA ADP |
| SSVC exploitation | none | CISA Vulnrichment |
| SSVC automatable | No | CISA Vulnrichment |
| SSVC technical impact | total | CISA Vulnrichment |
Affected versions (from the CVE record)#
- vllm-project vllm
>= 0.10.1, < 0.14.0(affected)
Description (from the CVE record)#
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.
Quoted from the CVE Program record.