Researchers developed AI software that detects paroxysmal atrial fibrillation (PAF) from ECG recordings taken during sinus rhythm, addressing a major diagnostic blind spot. The model was trained to recognize ECG signatures predictive of intermittent AF, potentially enabling earlier identification of patients at risk and informing monitoring or anticoagulation decisions. The study frames AI as a tool to extract latent diagnostic signals from routine ECGs and to expand screening beyond symptomatic episodes.