If you ask “what is this large language model vulnerable to?” the answer is: “yes.” You can prompt inject a chatbot almost any way you can think of. Prompt injection is easy because you can’t separ…