A type of "hack" where malicious instructions are hidden in an AI prompt to manipulate the behavior of an AI system, often overriding safety guidelines.
bro got his chatbot to ignore all the safety rules with a prompt injection and now it's saying whatever
+509