Abstract
The use of artificial intelligence (AI) in inborn errors of immunity offers transformative potential in diagnostics and disease management but faces multiple challenges that were discussed at the second Artificial Intelligence in Primary Immunodeficiency conference, held in New York City (March 19-22, 2025). The conference addressed 7 themes: predictive diagnostic algorithms, health equity, industry collaboration, advanced computational tools like large language models, patient-led AI initiatives, multiomics integration, and implementation science. Discussions highlighted the growing impact of AI on diagnostics, genomics, and health systems, emphasizing the need for high-quality, diverse datasets and ethical safeguards to ensure equitable application. Participants stressed that AI alone cannot resolve systemic inequities or delays in diagnosis. Challenges such as the lack of harmonized datasets, the complexity of integrating multiomics data, ethical concerns, and the difficulty of adapting solutions to low-resource settings were emphasized. Additionally, the use implementation science was pointed out as one of the major challenges to ensure applicability and scalability in real-world settings. This requires overcoming resistance to adoption, addressing infrastructure gaps, and ensuring regulatory compliance. Collaboration across academia, clinicians, patients, regulators, and industry is essential to ensure AI delivers equitable, lasting benefits for individuals with inborn errors of immunity.