Who helped read the slide? Patient disclosure in AI-assisted pathology

Artificial intelligence (AI) is increasingly embedded in diagnostic pathology, from biomarker quantification to tumour grading, yet patients are seldom aware when it contributes to their diagnosis. Whether, when and how this involvement should be disclosed remains unresolved. This review synthesises the legal, ethical and patient-preference literature on disclosure of AI use in pathology across the USA, the European Union and the UK. Current frameworks rarely mandate routine patient-level disclosure, favouring institutional governance and risk-proportionate transparency instead. Ethical analysis and patient-preference studies nonetheless suggest that disclosure matters most when AI is autonomous or materially shapes a high-stakes decision. Pathology poses a particular challenge: without a direct patient touchpoint, case-by-case disclosure becomes increasingly impractical as algorithmic contributions multiply. We propose a tiered, proportional framework (no patient-facing notice, notification or consent) matched to the AI system’s role and potential impact, with diagnostic accountability remaining with the pathologist and offer example report language for each tier. We conclude that pathology societies should develop and empirically evaluate standardised disclosure language before professional norms harden.