Artificial intelligence (AI) is rapidly reshaping the way medicines are discovered, developed, evaluated, and monitored. Recognizing both its transformative potential and its inherent risks, international regulators and standards bodies have jointly issued Guiding Principles of Good AI Practice in Drug Development to support the responsible use of AI across the entire drug product life cycle.
The guidance emphasizes that, despite technological advances, the fundamental basis for authorizing medicines remains unchanged: drugs must demonstrate quality, safety, and efficacy, with benefits clearly outweighing risks. AI should strengthen not replace these regulatory foundations, ensuring that innovation consistently serves patient safety and public health.
The document highlights the growing role of AI across nonclinical research, clinical trials, manufacturing, and post-marketing surveillance. When carefully managed, AI has the potential to accelerate development timelines, improve prediction of toxicity and efficacy, enhance pharmacovigilance, and reduce reliance on animal testing. However, the complexity of AI systems requires rigorous governance, transparency, and continuous oversight to ensure reliable and trustworthy outputs.
At the core of the guidance are ten principles designed to define good practice and create a common international foundation for AI in drug development.
First, AI systems must be human-centric by design, aligned with ethical values and focused on supporting clinical and regulatory decision-making rather than replacing it. A risk-based approach should guide validation, oversight, and mitigation strategies, proportionate to the system’s intended use and potential impact.
Strong adherence to legal, ethical, technical, and regulatory standards including Good Practices (GxP) is essential, supported by a clear and well-defined context of use for every AI application. Multidisciplinary expertise must be embedded throughout development, integrating both AI specialists and domain experts.
The guidance places particular emphasis on robust data governance and documentation, ensuring traceability of data sources, analytical decisions, and model development. Best practices in model design and software engineering should promote transparency, reliability, interpretability, and generalizability all critical for patient safety.
Performance must be assessed using risk-based validation frameworks that evaluate the full human–AI system, while life cycle management ensures continuous monitoring, periodic re-evaluation, and timely response to issues such as data drift. Finally, developers and regulators are urged to provide clear, accessible information to users and patients, explaining performance, limitations, and appropriate use in plain language.
Beyond technical guidance, the document underscores the importance of international collaboration. Harmonized standards, shared educational resources, and consensus frameworks are seen as essential to advancing responsible innovation and enabling regulators worldwide to keep pace with rapidly evolving AI technologies.
Together, these principles offer a practical roadmap for integrating AI into drug development in a way that protects patients, supports regulatory excellence, and unlocks the full potential of data-driven innovation.
Learn more: EuropaEMA and FDA set common principles for AI in medicine development | European Medicines Agency (EMA)
