The institutional logic behind medical AI deployment is straightforward and, from an equity standpoint, damaging: tools get built where the paying customers are. Commercializing AI in medical technology [1] means chasing reimbursable procedures and high-volume diagnostic pathways in health systems with the capital to integrate new software. The result is a research and deployment pipeline that generates advances in rare orbital disease detection and nuanced skin condition classification — genuinely useful tools — while the foundational bias problems documented in histopathology error research remain structurally unaddressed.