LiteLLM Vulnerability Chain Lets Low-Privilege Users Take Over AI Gateway Servers

Obsidian Security disclosed a critical chained vulnerability in LiteLLM allowing low‑privilege users to promote themselves to admin, exfiltrate stored API keys and model-provider secrets, and inject configurations that result in remote code execution on the gateway server; organizations running self‑hosted LiteLLM should patch immediately, audit accounts, rotate keys, restrict admin access, and review logs.