Popular AI Gateway LiteLLM Had Flaws That Let Low-Privilege Users Take Over Systems

Check Point Research has disclosed multiple high-impact vulnerabilities in LiteLLM (version 1.77.7), a widely used open-source LLM gateway. These gateways sit between business applications and AI model providers, and they often handle credentials, configuration and request routing. The flaws are tracked as CVE-2026-35029 and CVE-2026-59823, and they have been resolved as of LiteLLM version 1.83.9.

The most significant problem was a missing authorization check on the /config/update endpoint. According to the researchers, this allowed any authenticated user, including low-privileged accounts, to change the proxy's runtime configuration. That single weakness enabled arbitrary file read, administrative account takeover, and remote code execution through malicious Python modules loaded from cloud storage. A separate server-side request forgery (SSRF) flaw in the "Clientside LLM Credentials" feature let attackers redirect API calls and steal provider keys.

The researchers say the findings show that as AI infrastructure tools evolve quickly, basic security practices such as proper access control and input validation remain essential. Because a gateway can hold keys to several AI providers, a weakness in it can have wide consequences. This summary is based on the opening of the Check Point Research post, so businesses using LiteLLM should read the full report for further detail.