有馬 宏和

Approvals of artificial intelligence (AI)-based medical devices have been increasing worldwide. To avoid unnecessarily constraining AI innovation, regulatory authorities have introduced frameworks that promote good machine learning practices (GMLPs) and enable flexible management of postmarket changes. However, regulatory requirements are characterized by a dual structure combining medical device regulation and AI-specific governance, posing challenges to international harmonization. This study aims to identify universal regulatory requirements for AI-based medical device development. An integrated analytical framework derived from regulatory guidance and literature on ethical, legal, and social issues was updated using recent regulatory documents. We applied this framework to regulatory documents issued in Japan, the United States of America (USA), and the European Union (EU) and conducted a longitudinal comparative analysis across multiple time periods. The results show that coverage of evaluation requirements has improved over time, with most universal requirements addressed by at least one regulatory document in each region. At the same time, regions have increasingly emphasized distinct policy priorities, indicating a process of policy specialization. Japan has shifted its focus toward postmarket surveillance and data governance; the USA toward performance changes and risk-based change management, including predetermined change control plans; and the EU toward governance, conformity assessment, and fundamental rights protection. In contrast, cross-cutting issues such as human–machine teaming and fairness remain unevenly addressed. Based on these findings, this study presents practical options for developers, combining localization strategies aligned with regional priorities with longer-term universal approaches grounded in total product lifecycle-based evidence generation.