Particle.news

LiteLLM Vulnerabilities Let Low‑Privilege Users Seize Gateways and Steal AI Provider Keys

Public research shows missing authorization and an SSRF flaw can expose keys, run code, and give full admin control unless deployments are patched to v1.83.9 or later.

Overview

  • Check Point Research publicly released technical findings on October 13 that identify two high‑risk flaws in LiteLLM tracked as CVE‑2026‑35029 and CVE‑2026‑59823.
  • CVE‑2026‑35029 is a missing authorization check on the /config/update endpoint that allowed authenticated low‑privilege users to change runtime environment variables, read sensitive files, overwrite admin credentials, and trigger remote code execution by loading attacker modules from cloud storage.
  • CVE‑2026‑59823 is a server‑side request forgery in the Clientside LLM Credentials path that let a logged‑in user route requests to an attacker server and capture provider API keys that the gateway attaches to outbound requests.
  • LiteLLM maintainers published fixes in version 1.83.9 and researchers and vendors advise immediate upgrades, auditing gateway accounts and embedded instances, and rotating provider API keys if any sign of unauthorized access exists.
  • Because LiteLLM is widely used and often embedded inside other tools, this class of application‑security bug raises a large operational blast radius and follows earlier warnings about exposed password hashes and a March 2026 PyPI supply‑chain compromise.