Warning Over High-Risk Vulnerabilities in Widely Used LLM Gateway Tool
Check Point warns LiteLLM flaws CVE-2026-35029 and CVE-2026-59823 can give any user admin access and code execution.
Authenticated SSRF in LiteLLM Proxy via user_config
LiteLLM Proxy versions before 1.83.9 let an authenticated caller who already has a valid virtual key bypass request-body checks by nesting api_base inside user_config. The check is_request_body_safe blocked top-level api_base and base_url but did not inspect or reject that nested field. Because user_config builds the outbound router, the proxy then makes a server-side request to an internal or external host chosen by the caller, which can reveal endpoints the caller could not reach directly (CWE-918). Organizations running a LiteLLM AI gateway on those versions are affected. It is fixed in 1.83.9; it is not listed in CISA KEV and no public proof-of-concept is known.
What to do: Upgrade the LiteLLM Proxy to 1.83.9 or later. Until then, restrict who can hold virtual keys, limit the proxy's network reach to internal services, and review logs for user_config values that set api_base or similar outbound destinations.
| Product | Affected |
|---|---|
| BerriAI LiteLLM Proxy (AI Gateway) | prior to 1.83.9 |
Order-of-magnitude estimate by the model from install counts, market share and public scan data it knows; verify before quoting.
LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. Prior to 1.83.9, an authenticated LiteLLM Proxy caller with a valid virtual key can place api_base inside the user_config request body to bypass is_request_body_safe, which blocks top-level api_base and base_url but previously did not inspect or reject user_config. Because user_config constructs the outbound router, the nested destination redirects a server-side request to an internal or external host selected by the caller and can expose endpoints the caller cannot otherwise access. This issue is fixed in version 1.83.9.
Check Point warns LiteLLM flaws CVE-2026-35029 and CVE-2026-59823 can give any user admin access and code execution.
Check Point found patched LiteLLM flaws enabling config takeover, file read, RCE, and key theft.
CenterPoint says more than 1 million Texans were exposed, while Burger King Japan cites about 11.18 million members.
Today 1.6K readers · 2.5K views · 778 reads · all time 189K views