Artificial intelligence in rare diseases: toward clinical impact

Rare diseases (RDs) affect more than 400 million people worldwide, yet most patients remain undiagnosed or untreated due to delayed diagnosis and limited therapies. Artificial Intelligence (AI) offers powerful tools to address these unmet needs by integrating genomics, clinical, and imaging data to accelerate detection and therapeutic discovery. Nevertheless, most AI tools remain confined to proof of concept, exposing a persistent gap between algorithmic innovation and patient impact. Recent advances in generative models, federated learning (FL), and explainable AI (XAI) have begun to overcome barriers, such as data scarcity, privacy concerns, and biases. In this review, we highlight these developments and uniquely define the technical, ethical, and infrastructural priorities, including equity, Findable, Accessible, Interoperable, and Reusable (FAIR) data, and global coordination, required to translate AI for RDs into tangible clinical benefits.