Prompt injection, data leakage, and jailbreaking get lumped together under 'AI security,' but they require different defenses. Here's a Q&A breakdown of what enterprise teams actually need to implement.