Does AI in Healthcare Need an Editor-in-Chief? A Leading Example of True Human Oversight.

Abstract

The rapid integration of high-risk Artificial Intelligence (AI) into healthcare promises efficiency gains but also raises pressing questions about safety, accountability, and human agency. The European Union's AI Act classifies healthcare AI as high-risk and imposes strict human oversight requirements under Articles 14 and 26. Yet, translating this principle into practice exposes significant legal and operational ambiguity. The Act does not clearly identify who - among providers, deployers, or individual clinicians - bears the ongoing responsibility for oversight, risking fragmented accountability and potential harm through automation bias or misuse. This paper argues that effective oversight requires more than procedural compliance: it demands a systemic, institution-wide governance model. Analogous to an editor-in-chief coordinating an editorial team, hospitals must establish formal AI governance structures that ensure oversight is informed, continuous, and empowered. Deployers must create clear roles, policies, and training programs to equip human overseers with AI literacy and authority to challenge algorithmic outputs. Reconciling these duties with professional autonomy and regulatory constraints requires a shift from the narrow "Human-in-the-Loop" model toward a multidisciplinary AI Governance Committee. Such committees are essential to balance legal compliance with ethical responsibility, ensuring that AI enhances rather than undermines clinical judgment and patient advocacy.