Differences in perceptions of medical artificial intelligence between medical and non-medical professionals in Korea: a qualitative study

Purpose: Medical artificial intelligence (AI) is rapidly being integrated into clinical practice and healthcare systems, raising concerns regarding safety, accountability, and governance. Despite its increasing importance, empirical comparative studies examining differences in perceptions of medical AI among key expert groups remain limited. This study aimed to compare and analyze perceptions of medical AI among medical and non-medical professionals and to systematically identify commonalities and differences across policy-and governance-relevant domains. Methods: Focus group interviews using open-ended questions were conducted with 30 experts (15 medical and 15 non-medical professionals) who had direct experience with medical AI. Data were analyzed using inductive thematic analysis combined with qualitative comparative analysis. Analytical rigor was strengthened through independent coding and consensus-based discussions. Results: Both groups recognized the potential of medical AI to bring meaningful changes to healthcare systems. However, medical professionals primarily evaluated medical AI in terms of clinical applicability, patient safety, explainability, and accountability. In contrast, non-medical professionals emphasized technological maturity, scalability, data infrastructure, standardization, and system-integration potential. Group-specific patterns also emerged regarding perceived limitations, autonomy, educational priorities, and classification frameworks, particularly in relation to clinical risk management versus system-level design and governance considerations. Conclusion: Differences in perceptions of medical AI are systematically associated with distinct interpretive frames shaped by professional roles and responsibility structures. Effective implementation and policy design for medical AI therefore require an integrated approach that accounts for these structural differences. This study provides empirical evidence and a conceptual foundation for future quantitative and mixed-methods research on medical AI governance.