[chasd00]

"llm phishing" is a much better way to think about this than prompt injection. I'm going to start using that and your reasoning when trying to communicate this to staff in my company's security practice.