Legally mandated pre-deployment evaluations: promoting AI safety in clinical medicine across Africa
Date
2025
Authors
Gitau, M. L. M.
Journal Title
Journal ISSN
Volume Title
Publisher
Strathmore University
Abstract
The integration of advanced artificial intelligence (AI) into healthcare systems presents transformative potential for clinical medicine in Africa, yet it introduces profound context specific risks. This study identifies four primary risk vectors through content analysis. First, the potential for mistranslations and misinterpretations in diagnostic settings due to linguistic diversity; second, the inadequate integration of traditional medicine data, heightening the risk of adverse drug reactions and misguided therapeutic recommendations; third, genetic biases stemming from the underrepresentation of African populations, which may precipitate misdiagnoses; and fourth, the neglect of tropical diseases endemic to the region, further compounding diagnostic inaccuracies. Although dataset bias underpins all these concerns, each manifests in distinct ways that can compromise patient safety and clinical outcomes. In response to these challenges, the study advocates for urgent, legally grounded solutions. Through a comprehensive review of the literature and a qualitative analysis of current evaluation methods, legal instruments, and policy frameworks, the study argues that existing AI safety evaluation frameworks, predominantly developed in high-resource settings, are ill-suited to the African context as they fail to capture the distinctive risks inherent in the continent’s clinical environment. Instead, legally mandated, context-specific pre-deployment evaluations are essential. To operationalise this approach, a tripartite legal framework is proposed: (1) government involvement through data aggregation and verification of developer assessments; (2) private-sector incentives that promote rigorous, independent evaluations; and (3) enforceable developer obligations to adhere to context-aware safety standards.
Description
Full - text undergraduate research project
Keywords
Citation
Gitau, M. L. M. (2025). Legally mandated pre-deployment evaluations: Promoting AI safety in clinical medicine across Africa [Strathmore University]. http://hdl.handle.net/11071/16134