Genomic Data, Foundation Models, and the Governance of Biomedical AI

The convergence of genomics and advanced artificial intelligence is rapidly reshaping biomedical research and healthcare. The proliferation of large-scale molecular datasets, coupled with the emergence of powerful foundation models, has enabled increasingly sophisticated capabilities, including prediction of gene regulation, interpretation of genomic variants, identification of therapeutic targets, and stratification of disease risk. These developments are transforming workflows across basic research, translational science, and clinical practice.At the same time, the integration of genomic and multimodal biomedical data into AI systems raises complex ethical, legal, and societal challenges. Issues of privacy, data sharing, intellectual property, equity, and regulatory oversight are becoming more acute as models scale and move toward real-world deployment. Critically, many of these challenges are not external to the technology, but arise directly from decisions about data curation, model design, validation, and interpretation. Despite rapid technical progress, there remains limited understanding of how these experimental and computational practices should inform responsible governance of biomedical AI.This Research Topic aims to bring ethical, legal, and societal (ELSI) analysis into direct engagement with the technical development and application of AI systems in genomics and molecular medicine. Rather than treating governance as external to innovation, this collection focuses on how data practices, model architectures, evaluation standards, and translational pathways shape—and should be shaped by—ethical and regulatory considerations. We welcome contributions from computational biologists, genomic researchers, clinicians, and ELSI scholars whose work addresses the governance implications of real-world AI systems.We particularly encourage submissions that identify ELSI questions emerging from model development and deployment, including how training data, benchmark design, interpretability limits, and uncertainty affect clinical validity, accountability, and public trust. By connecting technical practice with governance frameworks, this Research Topic seeks to foster more grounded and actionable approaches to responsible biomedical AI.Submissions should address the intersections of AI innovation, genomic data, and governance, with a focus on challenges specific to large-scale molecular and multimodal datasets. Topics of interest include, but are not limited to:- Training data construction, curation, and bias in genomic AI systems- Reidentification risks and privacy limits in high-dimensional molecular data- Model interpretability, uncertainty, and clinical decision-making- Benchmarking, validation, and standards of evidence for biomedical AI- Variant interpretation and genotype–phenotype modeling in AI systems- Data sovereignty, cross-border data governance, and federated learning approaches- Intellectual property, model ownership, and benefit sharing in data-driven research- Transparency, reproducibility, and reporting standards in genomic AI- Equity, representation, and performance disparities across populations- Governance of multimodal foundation models integrating genomic, clinical, and imaging data Topic editors: Dov Greenbaum Source: https://www.frontiersin.org/research-topics/79723/genomic-data-foundation-models-and-the-governance-of-biomedical-ai