Scoping review of 60 sources in LMICs identifies barriers to AI adoption and governance gaps.

A scoping review systematically mapped 60 sources addressing AI governance, ethical, regulatory, or implementation issues within low- and middle-income countries (LMICs). The study aimed to characterize barriers to AI adoption and strategies for inclusive deployment across this diverse population. No randomized trials or comparative effectiveness data were available, as the design focused on literature synthesis rather than primary clinical outcomes.

Key results indicated that only 7.4% of LMICs have adopted national AI strategies. Furthermore, over 60% of AI models in these regions rely on non-representative datasets, a factor potentially increasing contextual bias. The distribution of study focus showed that 25 sources examined ethics, 17 addressed regulatory gaps, and 18 focused on implementation challenges. Additionally, fewer than 10% of institutions offer structured AI training, indicating a significant gap in workforce readiness.

Safety and tolerability data were not applicable to this observational mapping study, as no adverse events or discontinuations were reported. A critical limitation identified was the presence of substantial gaps in empirical research regarding the operationalization of AI in these settings. The evidence remains descriptive rather than causal, reflecting the early stage of research in this domain.

Practice relevance suggests that stakeholders should prioritize context-sensitive design and participatory governance to overcome identified barriers. Capacity building is essential to address the scarcity of structured training and the reliance on non-representative data. Clinicians and policymakers should interpret these findings as a call for improved infrastructure and ethical frameworks rather than immediate clinical guidelines.