An ecosystem approach to governing commercial actors in healthcare AI.

The ever-increasing attention to artificial intelligence applications in medicine and healthcare requires us to critically evaluate the effectiveness of the current U.S. regulatory environment in this arena. We outline a series of challenges facing the governance of healthcare AI, including an overreliance on self-regulation when many AI developers lack knowledge of relevant regulation or acknowledge their role in preventing harms, lack of shared responsibility for healthcare AI harms, and the lack of transparency and availability of evidence to assess healthcare AI's safety and effectiveness. We advocate for an ecosystem approach to developing potential solutions to these governance challenges. In particular, we identify actions that civil society organizations, funders, healthcare system purchasers, and payers can take to advance healthcare AI governance. We argue that these actors should make coordinated efforts toward advancing transparency and independent assessment of healthcare AI, and therefore can help fill gaps created by the emphasis on self-regulation.