CVE-2026-27893
High · 8.8CVSS
8.8
EPSS
1.8%
Published
Mar 27, 2026
CISA due date
—
The record
vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the user's explicit `--trust-remote-code=False` security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.
Affected packages
| Ecosystem | Package | Introduced | Fixed |
|---|
Related vulnerabilities
- CVE-2026-113730 OGNL is an Object Graph Navigation Library for Java. CWE-693
- CVE-2026-112077 A vulnerability was detected in radareorg radare2 6.2.0/6.2.2. CWE-693
- CVE-2026-88838 A flaw was found in BusyBox's mount utility. CWE-693
- CVE-2026-101899 Axios is a promise-based HTTP client for the browser and Node.js. CWE-693
- CVE-2026-107226 The AsyncHttpClient (AHC) library allows Java applications to easily execute HTTP… CWE-693
- CVE-2026-107697 FFmpeg before 8.1.3 contains a protection mechanism failure in the HLS demuxer that… CWE-693
- CVE-2026-105083 ImageMagick before 7.1.2-32 and 6.9.13-57 contains a policy bypass vulnerability in… CWE-693
- CVE-2026-101900 Axios is a promise-based HTTP client for the browser and Node.js. CWE-693