Medical AI Moves Beyond ‘Approval’ Toward Clinical Adoption as U.S. Clearances Top 1,500

An analysis finds that medical artificial intelligence (AI) is moving beyond “approval” and into a phase of adoption in real-world clinical settings.

Hwi-young Kim, director of the H-Data Strategy Center at Hallym University Chuncheon Sacred Heart Hospital, said in a report published on the 28th by the National Research Foundation of Korea’s Center for Bioeconomy Policy that “medical AI is moving beyond competition over performance in the laboratory and into a phase of integration into actual clinical workflows.”

According to an external tally, the U.S. Food and Drug Administration’s (FDA) public list of AI-enabled medical devices had surpassed 1,500 cumulative entries through decisions made in the first quarter of 2026. About three-quarters of them were concentrated in radiology.

However, the report noted that an increase in the number of cleared products is a separate matter from their sustained use in hospitals. It said bottlenecks to clinical adoption are shifting away from algorithm performance itself and toward clinical effectiveness, generalizability across institutions, integration with electronic medical records (EMRs) and picture archiving and communication systems (PACS), accountability and explainability, purchasing and reimbursement, and postmarket performance management.

The regulatory paradigm is also shifting from one-time clearance to total product life cycle (TPLC) management. The FDA has formalized predetermined change control plans (PCCPs) for AI software, while the International Medical Device Regulators Forum (IMDRF) is pursuing international harmonization of good machine learning practice (GMLP) and AI life cycle management.

South Korea established a separate legal framework for digital medical devices with the implementation of the Digital Medical Products Act in 2025. In 2026, it expanded a pathway for deferring new health technology assessments for medical devices with a high degree of independent use, using clinical evaluation data.

Kim recommended that the core of future policy should not be “rapid market entry” itself, but rather “linking conditional early market entry with the accumulation of real-world evidence (RWE), value-based reimbursement and postmarket management.” Clinical AI, he said, should be designed around a single life cycle encompassing clearance, evaluation, procurement, reimbursement, hospital information systems and monitoring.