Comment by spoaceman7777

The reality is the opposite of this post. LLMs are great at rapidly creating rough drafts, and humans are best (when properly trained) at critiquing LLM results.

So, LLMs are in fact better at blue-teaming, and humans are better at red-teaming.