These LLM models are just glorified parrots. Why are we even expecting them to display non-trivial insight about complex moral dilemmas? They're going to get things totally wrong and say a lot of silly stuff about the Nazis, that's just normal. No reason to act offended over it or demand 100% "Safety" from answers you might not like.