Global patterns and gaps in AI policies of national genomics initiatives

National genomics initiatives increasingly serve as resources for AI-driven research, yet governance of AI use within these programmes has received limited attention. We reviewed publicly available materials identified through internet-based searches across 240 countries and dependencies (conducted between March 2024 and February 2026), extracting initiative characteristics, AI use, and governance provisions. We identified 90 initiatives across 70 countries/territories. Current or planned AI use was identified in 32/90 initiatives (36%), spanning 29/70 countries/territories (41%), but publicly available AI-specific governance policies were identified in only three initiatives in two countries: All of Us, Genomics England, and UK Biobank. Where such policies existed, three governance strategies recurred: mandatory secure computation environments, treatment of trained models as potentially disclosive data outputs, and restrictions on external generative AI services. Most AI guidance appeared in FAQs or standalone documents rather than core data-use or participant-facing materials. Given the evolving legal landscape in Europe, including the EU AI Act and European Health Data Space, and globally, the choices these early adopters make are likely to set precedents beyond genomics.