Supportive AI in Digital Pathology Evidence, Limitations, and Governance for Latin America

Digital pathology has turned histology slides into analyzable data, but adoption of artificial intelligence in pathology remains uneven across institutions with differing access to scanners, storage, local validation, and trained staff. The reviewed evidence is most mature for bounded detection or prioritization tasks in whole-slide images and for support in grading or quantification. Inference of molecular alterations or tissue of origin remains primarily at the retrospective-development stage and should not be confused with prospective clinical utility. Generalization across scanners and laboratories, limited impact evaluations, and sparse regional coverage constrain implementation claims. Two direct Latin American sources describe barriers involving cost, connectivity, regulation, training, and interoperability, but they are expert-panel and narrative evidence rather than effectiveness evaluations. For Latin America, the responsible next step is to define reviewable tasks, validate each local configuration, and govern the lifecycle with explicit professional accountability. This document is a community evidence brief, not a systematic review or product evaluation.