Safeguarding generative AI. (Part 2)