Japan’s Pharmaceuticals and Medical Devices Agency (PMDA) has published a set of principles for the responsible use of artificial intelligence in regulatory activities. The document outlines a human-centred approach, risk-based governance, data management, oversight and ongoing evaluation of AI systems used in regulatory decision-making.

Scope of the principles

The document, Basic Principles for the Use of AI in Regulatory Activities (Early Consideration), was issued by PMDA’s Center for Regulatory Science on 2 November 2026. It addresses activities involving regulatory judgement and decision-making across the lifecycle of pharmaceuticals, quasi-drugs, cosmetics, medical devices and regenerative medical products.

The scope extends from non-clinical and clinical studies to manufacturing and post-marketing activities. Basic research and general medical practice are not its primary focus. The principles are intended for a broad group of stakeholders, including PMDA, applicants, marketing authorisation holders, manufacturers, sponsors, contract research organisations and developers or providers of AI systems. The development of AI systems used in these activities, including ongoing updates, may also fall within the scope of the document.

PMDA describes the publication as an early consideration document. It does not establish new regulatory requirements. The English version is a provisional translation; the Japanese original takes precedence in the event of any inconsistency.

A three-level framework

The principles are organised into three levels, beginning with an overarching human-centric principle, followed by six core principles and five principles for implementation and management.

Human-centric principle

AI is positioned as a tool to support human activities rather than replace human judgement. The interests of patients and the public remain the highest priority, while ultimate responsibility associated with the use of AI rests with humans.

Six core principles

The six core principles cover a risk-based approach; a clearly defined context of use and scope; data governance; fairness, accountability and transparency; multidisciplinary expertise; and compliance with applicable laws, regulations and standards.

Five implementation and management principles

The implementation principles address AI model and system design and development, evaluation, lifecycle management, human oversight and AI literacy. The document emphasises that evaluation should take account of the actual environment in which an AI system is used, rather than relying solely on technical performance measures.

Governance throughout the AI lifecycle

The framework highlights the need to manage the origin and characteristics of data, processing steps, change histories and traceability. It also emphasises the importance of continuous monitoring and periodic reviews after deployment, allowing performance and risks to be reassessed as the system, its data or its operating environment changes.

For pharmaceutical and medical product organisations, these principles provide a reference for considering how AI tools are selected, evaluated and overseen when they support regulatory activities. The emphasis is on defining the intended context of use, applying controls proportionate to risk, maintaining appropriate data governance and ensuring meaningful human oversight.

A reference document, not a new regulatory requirement

PMDA’s publication sets out an early framework for responsible AI use in regulatory activities. It should not be interpreted as a new GMP guideline or as introducing additional GMP requirements. The agency notes that the document may be reviewed and updated in light of technological developments, practical experience and international harmonisation.