CVE-2026-48746
CriticalCVSS 9.1Exploitation Probability (EPSS)
Elevated risk64th percentile - higher than 64% of all known CVEs
Summary
Vulnerability in vLLM versions 0.3.0 to 0.22.0 allows bypass of OpenAI API AuthenticationMiddleware. An attacker can use the API without providing the configured VLLM_API_KEY or --api-key.
Risk Assessment
An attacker can gain unauthorized access to the vLLM API, potentially leading to data leakage or abuse of computational resources.
Recommendation
Update vLLM to version 0.22.0 or later, which fixes this vulnerability.
Original NVD description (English source)
vLLM is an inference and serving engine for large language models (LLMs). From 0.3.0 until 0.22.0, a vulnerability in ASGI web servers and starlette's trust on those web servers enables an authentication bypass of the OpenAI API AuthenticationMiddleware. It allows to use the API without providing the configured VLLM_API_KEY or --api-key. This vulnerability is fixed in 0.22.0.