A Johns Hopkins team validated an AI-powered blood test designed to detect liver cancer across two geographically and biologically distinct patient populations. The findings, published July 31 in Cell Press Blue, reported the model’s ability not only to identify liver cancer but also to recover biological signals that the system relies on. The study strengthens the case for using genome-wide analysis of cell-free DNA (cfDNA) for multi-population detection, a key requirement for diagnostic tools intended for real-world deployment. By demonstrating performance outside a single cohort or reference population, the work addresses a major risk in many AI assays: generalizability. Further validation, including prospective clinical testing and calibration for confounders, will be needed before broad clinical adoption. Still, the multi-population readout supports continued investment in ctDNA/cfDNA AI-driven approaches for early detection and monitoring.