When LLM Gateways Go Wrong: Findings from LiteLLM

As cited

Copy frozen at 2026-10-13 18:15 UTC (site build).

vulnerabilities

When LLM Gateways Go Wrong: Findings from LiteLLM

A security assessment of LiteLLM, a widely deployed open-source language model gateway, identified multiple high-severity vulnerabilities tracked as CVE-2026-35029 and CVE-2026-59823. The flaws included missing role-based access control on the configuration update endpoint, allowing low-privileged users to modify proxy settings, inject malicious environment variables, and load arbitrary Python code for remote execution. A separate server-side request forgery vulnerability in the clientside credentials feature enabled attackers to redirect application programming interface (API) calls and extract provider keys.

Why it matters: Organizations deploying LiteLLM as a gateway for language model requests face compromise of stored credentials, administrative account takeover, and arbitrary code execution via an authenticated attacker with basic user privileges; upgrade to version 1.83.9 or later immediately to close these exposure paths.

Source published

2026-10-13 14:30 UTC

First seen by Cybersecurity Tracker

2026-10-13 17:21 UTC