Artificial Intelligence in U.S. Healthcare: Legal framework, liability, and ethical challenges with special focus on Pediatric Care

Abstract: This article maps the evolving U.S. regulatory environment for AI in healthcare and translates it into operational guidance with a pediatric lens. We distinguish ML-SaMD from CDS and identify lifecycle control points, including data curation, validation, predetermined change control plans, deployment, and monitoring. We then examine how the 21st Century Cures Act boundary for non-device CDS, FDA device pathways, HIPAA and HITECH privacy rules, and ONC HTI-1 transparency obligations interact in real clinical settings. Building on that analysis, the paper presents a responsibility matrix that allocates duties across manufacturers, institutions, clinicians, and EHR vendors, and proposes a pediatric safety framework that requires validation stratified by age, transparency that is usable in clinical practice, and drift monitoring linked to clear escalation protocols. We argue that existing legal frameworks only partially address algorithmic opacity, fairness, and the long-term outcome risks specific to children, and we offer policy and implementation recommendations to close these gaps. The result is actionable guidance for legal practitioners, health systems, and developers seeking to reconcile safety, equity, and continuous learning in AI-enabled care. --- Official reference version:Zenodo DOI: https://doi.org/10.5281/zenodo.17344639 Parallel version available:SSRN: https://ssrn.com/abstract=5468846 - DOI: https://doi.org/10.2139/ssrn.5468846