Exploitation ongoing with high activity in latest observed window (1 mentions)
Recommended action window: Immediate (within 24h)
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.
Priority
MEDIUM
Exploitation
ACTIVE
PoC
NONE
Patch
AVAILABLE
Momentum
STABLE
If you run products in this scope, you should treat this CVE as relevant to your environment.
| Date | Total | Labels |
|---|
| 2026-06-16 | 1 | Disclosure1 |
| 2026-06-22 | 2 | Disclosure1General1 |
| 2026-06-23 | 1 | Patch1 |
| 2026-06-24 | 1 | Active Exploitation1 |
Attention, elevated activities detected targeting vLLM (CVE-2026-48746) https://vuldb.com/vuln/372828/cti
Post summary
The post signals that elevated, potentially malicious activity has been observed against vLLM (CVE-2026-48746), indicating possible ongoing exploitation, but provides no proof of concept, exploit details, or mitigation information.
🚨 CRITICAL: CVE-2026-48746 (CVSS 9.1) - vLLM authentication bypass vulnerability affects versions 0.3.0-0.22.0. Attackers can access OpenAI API without API key. Patch to v0.22.0 immediately. #CVE #PatchNow #ThreatIntel https://t.co/x6e4drd5CG
Post summary
The tweet announces a critical authentication bypass flaw (CVE-2026‑48746) in vLLM, gives CVSS and version details, and urges an immediate patch to v0.22.0.
🚨*CVE* CVE-2026-48746 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 … https://www.cve.org/CVERecord?id=CVE-2026-48746 ----- Traducción: CVE-2026-48746 vLL… http://infoflow.cloud`
Post summary
The CVE-2026-48746 vulnerability affects vLLM's ASGI server integration up to version 0.22.0, with technical details available on the CVE record.
CVE-2026-48746 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 … https://www.cve.org/CVERecord?id=CVE-2026-48746
Post summary
The post briefly mentions CVE‑2026‑48746, noting its impact on vLLM 0.3.0–0.22.0 due to a flaw in ASGI servers and Starlette; it lacks details on exploits, patches, or active usage.
🚨 CRITICAL - Authentication Bypass in vLLM OpenAI API Middleware (CVE-2026-48746) vLLM's OpenAI API AuthenticationMiddleware can be bypassed due to improper trust in the reconstructed URL path from the ASGI scope. An attacker can manipulate the Host header to alter the parsed URL path used for authentication checks while maintaining access to protected routes through standard HTTP routing. This flaw allows unauthorized use of the API without an API key, posing a severe risk if the API is publicly exposed. 👉 Affected: vLLM >= 0.3.0, < 0.22.0 | fix 0.22.0
Post summary
vLLM's OpenAI API authentication bypass (CVE‑2026‑48746) permits unauthorized API use via Host header manipulation; no PoC or exploit code shared, but a patch (v0.22.0) is available.
1 of 1 entries
| Part | Vendor | Product | Version | Target SW | Target HW |
|---|---|---|---|---|---|
| App | vllm | vllm | - | - | - |