JMIR Publications

ABSTRACT

Background:

Trustworthy artificial intelligence (AI) in healthcare requires assurance frameworks that translate ethical principles into measurable governance and evaluation practices. However, few studies have compared how national frameworks operationalize these principles across the AI lifecycle.

Objective:

This study compares two contrasting national approaches to assuring trustworthy AI in healthcare: a consortium-driven and flexible model in the United States (Coalition for Health AI [CHAI] Responsible AI Guide) and a government-led and standardized model in South Korea (Trustworthy AI Guidelines).

Methods:

Using a seven-dimension rubric adapted from international assurance frameworks, seven independent evaluators scored each framework on a five-point scale (1 = absent, 5 = comprehensive) across core principles, lifecycle coverage, governance, stakeholder breadth, operational maturity, and public accessibility. Consensus was achieved through structured discussion

Results:

South Korea scored higher on lifecycle coverage, governance, and maturity, while CHAI scored higher on tools and accessibility, demonstrating complementary strengths that may inform future efforts to build interoperable assurance systems across jurisdictions

Conclusions:

The findings demonstrate how consortium-based and government-led frameworks can serve complementary roles in advancing globally harmonized and trustworthy AI practices in healthcare. By identifying points of convergence, this study provides a foundation for future efforts toward interoperable, cross-national assurance standards that enable responsible and scalable use of AI in clinical settings.