Vulnerable LiteLLM AI Gateways Expose Root Access and Cloud Credentials

• •9.6% of LiteLLM AI gateways are vulnerable to root access and credential theft.
• •Attackers can exploit default master keys or lack of authentication to gain access.
• •The vulnerabilities pose a serious risk for organizations with internet-facing deployments.

A significant vulnerability has been identified in LiteLLM AI gateways, with nearly 10% of internet-exposed systems accepting a default master key or requiring no authentication. This flaw allows attackers to execute root-level code and steal sensitive cloud credentials, posing a serious risk to organizations using these gateways. An internet scan revealed that 294 out of 3,074 publicly reachable LiteLLM instances were vulnerable. The weaknesses could lead to LLMjacking and unauthorized access to connected tools and environments. Organizations with internet-facing LiteLLM deployments are particularly at risk. The vulnerabilities have been disclosed recently, prompting urgent advisories for affected users to secure their systems.