I don’t understand why they don’t use a second model to detect falsehoods instead of trying to fix it in the original LLM?

I wonder what the impact on piracy will be? It’s probably harder to rip streaming services than it is to rip blu-rays…

As an AI model, I cannot answer duplicate questions. Please check this completely unrelated question instead.