How LLM Chatbots Get Hacked: Prompt Injection, Prompt Leakage, and Model Fingerprinting

LLM chatbots fail in a different way than normal apps. A normal app receives input, checks permissions, calls APIs, and returns a...