Rigor and transparency in FDA authorization of AI-enabled medical devices: a critical distinction

Recent analyses have highlighted important concerns regarding the extent of clinical validation conducted for FDA-authorized AI devices. Here, we draw a distinction between the rigor of validation performed for FDA authorization and the transparency into this validation. We argue that these concepts are often entangled, leading to inferences about rigor that may sometimes instead reflect limited transparency. We first clarify this distinction based on FDA guidelines and unwritten rules learned from our experience developing multiple FDA-cleared AI devices. We then propose key considerations for enhancing both aspects as AI applications proliferate. Altogether, we emphasize the need for a nuanced view of regulatory practices to guide actionable frameworks that balance risk and innovation in healthcare AI.