Overview
- Security researcher Check Point disclosed mid-October 2026 two high-risk vulnerabilities in LiteLLM that let authenticated low‑privilege users gain administrator control, read server files, execute arbitrary code, and harvest provider API keys.
- The first bug, CVE-2026-35029, is a missing authorization check on the /config/update endpoint that lets attackers overwrite environment variables to extract files, replace admin credentials, or register pass-through endpoints that load and run attacker code.
- The second bug, CVE-2026-59823, is a server-side request forgery in the Clientside LLM Credentials flow that can route /chat/completions requests to an attacker’s server and expose provider API keys sent in request headers.
- LiteLLM shipped fixes in version 1.83.9 and organizations are urged to upgrade immediately, audit gateway accounts and roles, and rotate any AI provider API keys if unauthorized access is suspected.
- Because LiteLLM is widely embedded in tools and has very large download counts, the report warns that gateway compromise is a high-value single point of failure and it reinforces the need to include LLM gateways in regular security assessments and supply-chain checks.