JMIR Publications

ABSTRACT

Clinical AI governance has expanded alongside clinical AI adoption, yet most existing frameworks remain reactive, focusing on near-term validation and retrospective risk detection rather than on the iterative, adaptive, and multi-actor nature of contemporary AI deployment. This mismatch is becoming increasingly consequential as AI systems grow more complex, autonomous, and embedded in care relationships that existing oversight mechanisms were not designed to govern. In this viewpoint, we introduce the futures framework for clinical AI governance (FF-CAIG), an anticipatory governance framework grounded in three futures methodologies (the three horizons model, scenario planning, and causal layered analysis) and operationalized through an emerging clinical AI risk taxonomy that links insights from these methods to actionable governance domains. Applied across near-term, transitional, and longer-term governance horizons, FF-CAIG identifies cross-horizon priorities, including stronger predeployment equity evaluation, clearer lifecycle accountability, clinician AI oversight competencies, and governance models for increasingly autonomous systems. The three horizons model structures governance challenges across temporal phases, scenario stress-testing identifies robust priorities under regulatory uncertainty, and causal layered analysis surfaces the structural assumptions and cultural narratives that shape which risks are recognized and whose interests are centered in governance design. We illustrate FF-CAIG through three representative clinical AI deployment patterns and discuss its limitations, including differential compliance burdens, risks of overdocumentation, and variable feasibility across jurisdictions. FF-CAIG is intended not as a prescriptive policy instrument but as a structured tool for regulators, health system leaders, developers, and researchers seeking more prospective and systems-oriented approaches to clinical AI oversight.