The first time someone fell for a phishing email that was LLM generated we had proof of AI telling convincing lies.