Ethical aspects of artificial intelligence use in neonatal intensive care units: a scoping review

This study aims to map ethical, legal, social, professional, and implementation issues associated with artificial intelligence (AI) in neonatal intensive care units (NICUs). A JBI-informed scoping review using the population-concept-context framework was reported according to PRISMA-ScR. Peer-reviewed English-language articles published from 1 January 2016 to 31 May 2026 were eligible. Because the original exports and screening log were unavailable, a documented updated rerun was completed on 13 July 2026 using public PubMed-indexed bibliographic searching, supplementary publisher and bibliographic web searching, and backward and forward citation chaining. Exact strategies and record-level decisions are provided as online resources. The retrieval log contained 78 record captures. After removal of 21 duplicates, 57 unique records were screened; 45 full texts were assessed, 14 were excluded with documented reasons, and 31 sources were included. Eight recurring domains were identified: data governance; bias and fairness; transparency and explainability; human oversight; accountability and surveillance; parental engagement; professional readiness and workflow; and equitable implementation. Empirical evidence was concentrated on parent and nurse perceptions, pain assessment, counseling, and explainability, whereas consent processes, subgroup fairness, liability, and post-deployment safety remained under-studied.

Ethically responsible NICU AI requires secure governance, local and subgroup validation, understandable communication, active clinician oversight, defined accountability, staff and family engagement, and prospective monitoring. Generative AI should remain supervised and should not replace clinician-family communication.

• Artificial intelligence is increasingly being developed for neonatal outcome prediction, monitoring, pain assessment, clinical decision support, documentation, and family communication, but relatively few systems have progressed to validated routine NICU use. • The use of AI in neonatal care raises concerns regarding privacy, algorithmic bias, explainability, accountability, human oversight, parental trust, and equitable access.

• This scoping review identifies eight recurring ethical and implementation domains for NICU AI: data governance, fairness, transparency, human oversight, accountability, parental engagement, professional readiness, and equitable implementation. • Current empirical evidence is concentrated on stakeholder perceptions, pain assessment, counseling, and explainability, while consent processes, subgroup fairness, legal responsibility, and post-deployment safety monitoring remain insufficiently studied.