Human in the loop requirement and AI healthcare applications in low-resource settings: A narrative review.

Background. Artificial intelligence (AI) applications in healthcare provision have the potential to universalise access to the right to health, particularly in low-resource settings such as rural and remote regions in which AI is deployed to fill in medical expertise gaps. However, a dominant theme in evolving regulatory approaches is the human in the loop (HITL) requirement in AI healthcare applications to ensure the safety and protection of human rights. Objective. To review HITL requirements in AI healthcare applications and inform how best to regulate AI applications in low-resource settings. Method. We conducted a narrative review on HITL requirements in AI healthcare applications to assess its practicality in low-resource settings. Results. HITL requirements in low-resource settings are impractical as AI applications are deployed to fill in gaps of insufficient medical experts. Conclusion. There is a need for a shift in regulatory approaches from primarily risk-based to an approach that supports the accessibility of AI healthcare applications in low-resource settings. An approach anchored on the human right to science ensures both the safety requirements and access to the benefits of AI systems in healthcare provision. [ABSTRACT FROM AUTHOR]