Fresh LiteLLM Vulnerability Exploited Shortly After Disclosure

A critical-severity vulnerability in the open-source AI gateway LiteLLM was exploited shortly after its public disclosure, according to a report from cybersecurity firm Sysdig. The exploit allowed attackers to access database tables containing sensitive information. The vulnerability was disclosed prior to the exploitation, and attackers acted within days to take advantage of it.

Sysdig documented the exploit, noting that it targeted LiteLLM, a tool used by many organizations to manage access to large language models (LLMs). The exact nature of the sensitive data accessed was not fully detailed, but such databases often contain API keys, user prompts, and configuration data. The speed of exploitation highlights the risks associated with widely used open-source components in AI infrastructure.

As adoption of tools like LiteLLM increases, the window for patching vulnerabilities after disclosure narrows. Organizations relying on LiteLLM are advised to apply patches promptly and review their security configurations.