Reconceptualizing Health Data Governance in the Era of Predictive AI and Continuous Multi-Omics

The convergence of high-throughput multi-omics, continuous patient-generated health data, and predictive artificial intelligence (AI) has strained the legal and ethical vocabularies historically used to describe rights over health information. This narrative review examines why classical personal-property and intellectual-property frameworks sit uneasily with contemporary biomedical data, situates the resulting governance gap within a documented case of asymmetric value capture (the 2025 bankruptcy sale of 23andMe), and evaluates three governance responses, dynamic consent, fiduciary data trusts, and statutory data spaces, principally the European Health Data Space (EHDS), against a comparative snapshot of the United States, United Kingdom, China, and India as of September 2026. Rather than treating data-protection rights as equivalent to ownership, the review disaggregates the objects in play (biospecimens, raw measurements, curated datasets, derived data, trained models, and their outputs) and proposes an operational, accountability-mapped stewardship framework in place of the binary language of ownership. A closing section situates the debate historically, arguing that United States patent law already answered a narrower version of “who owns the genome”, by holding isolated genomic DNA unpatentable while leaving synthetic derivatives and downstream algorithms fully capturable, and that this asymmetry, mirrored in the sectoral gap between HIPAA and the more general but historically fragmented European approach, is the deeper structural problem the stewardship model must address. This is a narrative rather than systematic review; its method, scope, and limitations are set out in §2.