A LLM-based system cannot produce results that it hasn’t explicitly been trained on, and even making its best approximation with given data will never give results based on the real thing. That, and most of the crap that LLMs “”“censor”“” are legal self-defense
Add ground beef and worcestershire