High-Risk Vulnerabilities Discovered in LiteLLM Gateway
Researchers identified flaws allowing unauthorized administrative access and traffic redirection in the popular AI tool.
Updated on Oct. 13, 2026 in Artificial Intelligence
Live Poll
Do you trust open-source software to keep your business data and digital operations secure?
Check Point Research has identified two high-risk vulnerabilities in LiteLLM, an open-source gateway used to connect applications to AI providers. These flaws, tracked as CVE-2026-35029 and CVE-2026-59823, could allow unauthorized parties to gain administrative control or redirect traffic.
Why it matters
The vulnerabilities impact an essential piece of infrastructure that bridges internal enterprise and startup environments with external AI models. Given LiteLLM's widespread adoption, these flaws represent a significant security surface for organizations integrating LLMs into their workflows.
LiteLLM serves as a high-traffic gateway with 97 million monthly downloads, connecting applications to AI providers. The discovered vulnerabilities allowed read-only users to elevate privileges to administrative control or manipulate data flows through traffic redirection.
The players
Check Point Research
A threat intelligence unit that monitors global software vulnerabilities and infrastructure risks.
LiteLLM
An open-source gateway software package that manages API connections between internal enterprise applications and external AI model providers.
The details
LiteLLM acts as an open-source gateway, a software layer that standardizes requests between diverse applications and various large language model (LLM) providers. CVE-2026-35029 specifically allowed read-only users to bypass authentication checks to access administrative functions and hijack the gateway. Meanwhile, CVE-2026-59823 created a pathway for attackers to intercept and redirect outbound traffic, potentially exposing sensitive model prompts or internal data.
Timeline
October 13, 2026: Discovery of the vulnerabilities reported.
The Tech Race
As organizations rush to integrate LLMs, the infrastructure layer connecting these models is becoming a primary target for security researchers. This discovery highlights the growing pressure on widely adopted open-source gateways to match the hardened security standards of proprietary enterprise software.
Organizations currently running LiteLLM deployments in their production or development environments should prioritize auditing their gateway logs for unauthorized traffic. Security teams should ensure their versions are updated to mitigate the risks associated with the reported CVE-2026-35029 and CVE-2026-59823 flaws.
The takeaway
The ubiquity of LiteLLM in the AI stack makes it a high-value target for researchers and attackers alike. IT administrators should verify their current deployment versions and monitor official security advisories for patches addressing these specific CVE entries.
Further reading
For broader context on how organizations secure their LLM pipelines, visit our coverage on Artificial Intelligence.
Live Poll
Do you trust open-source software to keep your business data and digital operations secure?
