Two High-Risk Flaws in Popular AI Gateway LiteLLM Could Hand Attackers Full Control

Researchers at Check Point have uncovered two high-risk vulnerabilities in LiteLLM, an open-source LLM gateway with almost 100,000 downloads a month. Gateways like this sit between a business's applications and AI providers such as OpenAI, Google and Anthropic, forwarding requests to the right model. The flaws, tracked as CVE-2026-35029 and CVE-2026-59823, affect LiteLLM versions older than 1.83.9, and were detailed in a Check Point blog post published on October 13.

The risk is significant because gateways often centralise credential management, logging, usage tracking and user permissions. That means they store API keys for every AI provider an organisation uses. According to the researchers, CVE-2026-35029 involves a missing authorisation check that could let an attacker change the gateway's runtime configuration, potentially leading to full system takeover and remote code execution. CVE-2026-59823 involves server-side request forgery, which could redirect the gateway's outbound requests and expose AI provider API keys.

Together, the flaws could allow any user to gain administrative access, with file access, credential theft and remote code execution possible. As Check Point's Mateo Rojas-Carulla put it, "anyone holding an account on a vulnerable LiteLLM deployment could take over the gateway and use it for malicious purposes." This summary is based on the opening of the original article, so further details may be available in the full report.