Prompt injection and jailbreaking are two different types of attacks with different implications and risks. Prompt injection is a class of attacks against applications built on Large Language Models (LLMs) that involve concatenating untrusted user input with a trusted prompt. Jailbreaking, on the other hand, is a class of attacks that attempt to subvert safety filters built into LLMs themselves. The risks of prompt injection are more serious as they target applications built on the models and can lead to unauthorized access to confidential data and privileged tools.
4 Impressions