Research Ethics Committees, Community Engagement, and Artificial Intelligence: A Systematic Review of Readiness, Capacity, and Governance in Africa. (Preprint)

BACKGROUND Background: Africa bears over 20% of the global disease burden but contributes only 3% of clinical trials. Artificial intelligence (AI) is increasingly used in trial design, recruitment, monitoring, and analysis, yet the readiness of African Research Ethics Committees (RECs) and community engagement (CE) frameworks for AI integrated research remain unknown. OBJECTIVE Objective: To systematically identify, appraise, and synthesize empirical evidence and policy guidance on (1) REC capacity to review AI enabled clinical trials in Africa, (2) documented challenges for ethical oversight, and (3) community engagement strategies that address AI specific issues. METHODS Methods: The study was a systematic review conducted using the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. Empirical peer-review articles on formal policy analysis on AI in clinical trials, REC function/ethics review, or community engagement in African settings published between January 2000 and April 2026 in English or French, were retrieved from seven databases: PubMed/MEDLINE, Web of Science, SCOPUS, Cochrane Library, CINAHL, African Journals Online (AJOL), and Google Scholar as well as grey literature from the WHO, African Union, EDCTP, and national REC websites. Two reviewers independently screened, extracted data, and assessed quality using the Mixed Methods Appraisal Tool (MMAT) version 2018. Synthesis was thematic with harvest plots for capacity indicators. RESULTS Results: Of 1,847 screened records, 33 met the inclusion criteria, with only 24% (n=8) specifically addressing the intersection of AI and Research Ethics Committees (RECs). Findings reveal critical institutional deficits: no identified African REC utilized dedicated AI protocols, specialized training, or technical advisors. Primary challenges included algorithmic opacity (82%), data extractivism (73%), and accountability gaps (67%), while AI-adapted community engagement remained rare and unevaluated (n=6). Given the predominance of descriptive data and lack of intervention evaluations, the evidence base warrants only moderate-to-low confidence. CONCLUSIONS Conclusions: There is a critical absence of operational REC readiness for AI integrated trials in Africa. Community engagement remains conceptual rather than implemented. Immediate priorities include mandatory AI literacy training for REC members, development of African led algorithmic audit frameworks, and pilot testing of participatory CE models with embedded evaluation.