🚨 Lab: Exploiting LLM APIs with Excessive Agency (Apprentice)
Press enter or click to view image in full size
Non-member access: link
🎯 Objective
Use the LLM-powered assistant to delete the user carlos.
🧩 Introduction
Modern applications are increasingly integrating Large Language Models (LLMs) with backend systems. While powerful, this introduces a dangerous risk known as:
Excessive Agency — where an LLM has access to sensitive APIs and can be manipulated into using them unsafely.
In this lab, the LLM is connected to internal APIs capable of performing user management operations, including deleting users.
🏗️ Application Behavior
The chatbot acts as an assistant that can:
- Manage users
- Perform administrative operations
- Interact with backend APIs
⚠️ The critical flaw:
The backend blindly trusts LLM-generated actions.