Remote Code Execution via Prompt Injection in vanna-ai/vanna

CVE-2024-5826

What is CVE-2024-5826?

The Vanna AI product experiences a vulnerability allowing for remote code execution due to prompt injection within the vanna.ask function. This issue stems from an absence of sandboxing when executing code generated by language models. As a result, untrusted input can manipulate the code executed through the exec function in src/vanna/base/base.py. An attacker can exploit this vulnerability to achieve remote code execution on the backend server of the application, which may lead to full server compromise.

References