PMDA outlines data governance for AI used in regulatory activities
PMDA's 2 November early consideration addresses data provenance, performance assessment and human oversight; it creates no new legal requirement.
Japan's Pharmaceuticals and Medical Devices Agency published 'Basic Considerations on the Use of AI in Regulatory Activities (Early Consideration)' on 2 November 2026. It addresses AI used to support regulatory judgments and decisions across the lifecycle of medicines, quasi-drugs, cosmetics, medical devices and regenerative medical products. Basic research is not its main scope. This is reference guidance, not a new binding rule; it sets no enforcement date or transition period.
The document calls for a defined purpose and scope, risk-based data and performance assessment, traceable records of data characteristics, sources, processing, changes and decision rationale, validation in the intended use environment, ongoing monitoring and human oversight. People retain final responsibility for decisions.
For pharmaceutical laboratories whose AI-assisted interpretation of analytical data feeds into regulatory decisions, the document offers a basis to review input provenance, processing records, fit-for-purpose performance and human review. This is an application of the agency's principles, not a laboratory-specific mandate.