Actively exploitedCVSS 8.2 · HIGH

CVE-2026-59822Authentication bypass in LiteLLM

BerriAI LiteLLM contains an improper authentication vulnerability in the MCP Streamable HTTP endpoint that could allow an unauthenticated attacker to establish an authenticated MCP session using an arbitrary Bearer token.

View on NVD

Analysis

LiteLLM, el popular proxy para APIs de inteligencia artificial, presenta una vulnerabilidad que permite a atacantes saltar la validación de llaves de API mediante un encabezado de autorización manipulado. El fallo permite el acceso no autorizado a herramientas MCP (Model Context Protocol) y está siendo explotado activamente en aplicaciones reales. Se recomienda actualizar a la versión 1.84.0 de forma inmediata.

Relevant roles

PythonaiIABackendMachineLearningDataScienceciberseguridad

Severity

Score: 8.2(HIGH)

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

AV: NETWORK

AC: LOW

PR: NONE

UI: NONE

S: UNCHANGED

C: HIGH

I: LOW

A: NONE

Weakness (CWE): CWE-287CWE-306

CISA KEV

Added to KEV: 2026-09-02

Federal patch deadline: 2026-09-16

Known ransomware use: Unknown

Required action

Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.

EPSS

Probability of exploitation (next 30 days): 0.0052 (0.5%)

Percentile: 41.9%

EPSS: 2026-09-02

Affects

litellm:litellm

Technical description

LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. Prior to 1.84.0, LiteLLM's MCP Streamable HTTP endpoint allowed an unauthenticated attacker to use a fabricated Authorization header to trigger an OAuth2 passthrough fallback path that replaced failed LiteLLM key validation with an empty UserAPIKeyAuth() object, allowing requests to reach MCP tooling without a valid LiteLLM key. This issue is fixed in version 1.84.0.

Published: 7/8/2026, 8:16:57 PM

Last modified: 9/2/2026, 8:16:18 PM

References

How to read this card

Actively exploited

KEV (Known Exploited Vulnerabilities) is CISA's catalog of CVEs confirmed as actively exploited in the real world. If a CVE is on KEV, someone is already using it in attacks.

CVSS 10.0

CVSS 10.0 — máxima severidad teórica. Explotable de forma remota, sin privilegios, con impacto total. Aún no confirmada como explotada en el mundo real, pero parchea ya. Solo software de uso masivo: un 10.0 en un plugin de nicho no llega aquí.

AI-recommended

Un modelo de lenguaje revisó el CVE y consideró que afecta a software ampliamente desplegado o herramientas comunes de desarrollo. Un humano del equipo aprobó la publicación antes de que llegara aquí.

Categorías de impacto que evalúa la IA

widely_deployed_infra — kernel Linux, OpenSSL, glibc, k8s, Docker, nginx, OpenSSH.

popular_dev_tooling — Node.js, Python, Go, React, Next.js, librerías comunes de npm/pip.

common_consumer_software — Chrome, Firefox, Windows, macOS, apps móviles populares.

enterprise_saas — GitHub, GitLab, Atlassian, Salesforce, Cisco, Fortinet, Ivanti.

niche_software — CMS regionales, herramientas de hobby, librerías de poco uso.

not_relevant — firmware específico de proveedor, software descontinuado.

EPSS — EPSS (Exploit Prediction Scoring System) by FIRST.org estimates the probability that a CVE will be exploited in the next 30 days. Range is 0 to 1; very fresh CVEs typically score low until evidence accumulates.