CVE-2025-66448: vLLM remote code execution vulnerability in config loading

# CVE-2025-66448: vLLM remote code execution vulnerability in config loading

https://gcdradar.vercel.app/incidents/cve-2025-66448-vllm-remote-code-execution-vulnerability-in-c · AIR-2025-0381 · closed 2025-12-01

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2025-12-01 (happened 2025-12-02)
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2025-66448
- CVSS: 7.1

## Summary

vLLM is vulnerable to remote code execution through the transformers_utils configuration loader. The flaw allows executing remote Python code when loading model configurations.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-12-02: CVE-2025-66448: vLLM vulnerable to remote code execution via transformers_utils/get_config (https://github.com/advisories/GHSA-8fr4-5q9j-m8gm)
- 2025-12-01: CVE-2025-66448: vLLM is an inference and serving engine for large language models (LLMs). (https://nvd.nist.gov/vuln/detail/CVE-2025-66448)

## Sources

- 2025-12-01 NVD: CVE-2025-66448: vLLM is an inference and serving engine for large language models (LLMs). https://nvd.nist.gov/vuln/detail/CVE-2025-66448
- 2025-12-02 GitHub Advisory Database: CVE-2025-66448: vLLM vulnerable to remote code execution via transformers_utils/get_config https://github.com/advisories/GHSA-8fr4-5q9j-m8gm

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.