Access Gaps, Algorithmic Bias, and Ethical Challenges in AI-Enabled Genomic Personalized Medicine: Equity and Governance in 2026
By 2026, artificial intelligence (AI) and clinical genomics have converged into a working (if uneven) operational reality. Whole-genome sequencing (WGS) now costs under USD 500 in well-equipped U.S. and Chinese laboratories, foundation models such as AlphaMissense, scGPT, and Nicheformer are reshaping variant interpretation and multi-omic analysis, and the U.S. Food and Drug Administration (FDA) had authorized more than 1,200 AI-enabled medical devices by July 2025, with cumulative totals reaching approximately 1,451 by year-end, 1,524 by late March 2026, and 1,614 by the close of the second quarter of 2026. Radiology accounts for roughly three quarters of that cumulative list, and as of September 2026 no authorized device used generative AI or a large language model, which places the foundation models discussed below outside the cleared-device pathway rather than inside it. National programs, from the U.S. All of Us cohort (more than 535,000 whole genome sequences released to researchers as of June 2026) to the UK Generation Study (100,000 newborn genomes) and GenomeIndia (10,074 genomes from 83 populations), are pushing AI-enabled personalized medicine out of the academic medical center and into routine care. Yet the same year also saw the 23andMe bankruptcy, the sale of substantially all 23andMe assets to TTAM Research Institute for USD 305 million (closed 14 July 2025), and a £2.31 million UK Information Commissioner's Office (ICO) fine; the persistence of ancestry-driven performance gaps in polygenic risk scores (PRS) on the order of 4–8 percentage points of area under the receiver operating characteristic curve (AUC); and continued under-representation of African, Indigenous, and Pacific populations in the reference datasets that train AI models. This paper synthesizes peer-reviewed literature, regulatory guidance, and 2024–2026 policy documents through a five-gap analytical taxonomy (data representation, performance and validation, access and infrastructure, regulatory and governance, and ethical/privacy/social) to argue that the default trajectory of AI-genomic medicine is on track to widen, not narrow, global health disparities unless equity and governance are treated as core design requirements. It outlines the technical terrain, documents access and bias gaps, examines ethical and legal shortfalls, and offers concrete, stakeholder-mapped recommendations for developers, regulators, health systems, and researchers.
Research library