CVE-2026-48746

Medium

Elevated severity or exploit probability.

CVSS base

9.1 CRITICAL

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:H

EPSS — probability of exploitation (30 days)

1.2%

65.6th percentile

CISA KEV

Not listed

Weakness / dates

CWE-501

Published 2026-06-22 · modified 2026-09-16

CVSS breakdown

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:H

Attack Vector	N	Network
Attack Complexity	L	Low
Privileges Required	N	None
User Interaction	N	None
Scope	U	Unchanged
Confidentiality	H	High
Integrity	N	None
Availability	H	High

Timeline

2026-06-22 — Published (NVD)
2026-09-16 — Last modified (NVD)

Description

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.

Affected

vllm

References

Official: NVD · CVE.org