Ethics in AI Research: Emerging Governance Challenges for Institutional Review Boards in Health Systems

Abstract

Abstract Artificial intelligence (AI) is increasingly integrated into health research, creating ethical questions that extend beyond privacy, consent, bias, transparency, accountability, and safety. Existing bioethical principles and AI governance frameworks provide essential foundations but complex health-data ecosystems also raise institutional questions concerning reviewer capability, stewardship, reciprocity, authority, and downstream value. This Perspective proposes a stewardship-centered extension organized around four non-exhaustive governance domains: (1) IRB capability asymmetry; (2) data stewardship and sovereignty; (3) data contribution justice; and (4) authority, ownership, and intellectual property. The framework distinguishes ethical relevance from governance authority. IRBs should not adjudicate contracts, intellectual property allocation, national data policy, or technical matters outside their mandate. Instead, a recognize-assess-refer approach is proposed: IRBs recognize governance issues that may affect participant protection, justice, scientific integrity, or the ethical acceptability of research; assess implications within their mandate and available expertise; and refer matters requiring separate authority or specialist expertise to the appropriate institutional or regulatory body. The article further situates consent within layered governance, distinguishing participant consent, ethical approval, institutional authorization, data-use arrangements, and downstream rights. The framework is conceptual rather than an empirically validated standard and is intended to support coordination across research ethics and institutional governance as AI research becomes increasingly data-intensive, multidisciplinary, commercial, and cross-border.