EMA and FDA align on principles for use of AI across the medicines lifecycle 

The European Medicines Agency (EMA) and the US Food and Drug Administration (FDA) have jointly published ‘Guiding principles of good AI practice in drug development’, setting out ten high-level principles to guide the responsible use of AI across the medicines lifecycle. 

Developed through collaboration between FDA’s Center for Drug Evaluation and Research (CDER) and Center for Biologics Evaluation and Research (CBER), the principles are intended to support innovation while reinforcing patient safety and regulatory decision-making. 

The guidance reflects the growing use of AI across nonclinical research, clinical development, manufacturing, post-marketing surveillance, and pharmacovigilance. Both regulators recognize AI’s potential to analyze complex datasets, improve predictions of human toxicity and efficacy, shorten development timelines, and support more effective regulatory oversight. As the joint EMA–FDA document notes, AI technologies are anticipated to: 

“Support a multi-faceted approach that promotes innovation, reduces time-to-market, strengthens regulatory excellence and pharmacovigilance, and decreases reliance on animal testing by improving the prediction of toxicity and efficacy in humans.”  

At the same time, the agencies stress that medicines must continue to meet established standards of quality, safety, and efficacy, and that AI should strengthen, not undermine, these requirements. The principles therefore emphasize lifecycle management, proportionality, and human oversight, stating that AI use should “align with ethical and human-centric values” and support expert judgement. The ten principles are intended to be applied together and across all phases of medicines development:  

1. Human-centric by design: AI systems should align with ethical, human-centric values and be designed to support, not replace, human judgement. 
2. Risk-based approach: Validation, oversight, and risk-mitigation activities should be proportionate to the AI system’s context of use and the level of model risk. 
3. Adherence to standards: Compliance with applicable legal, ethical, scientific, technical, cybersecurity, and regulatory standards, including Good Practice (GxP) requirements, is a baseline expectation. 
4. Clear context of use: The role, scope, and purpose of each AI system should be explicitly defined so that outputs can be interpreted and applied appropriately. 
5. Multidisciplinary expertise: Effective development and deployment require integrated expertise across AI development, clinical and nonclinical science, regulation, and methodology throughout the system lifecycle. 
6. Data governance and documentation: Regulators emphasize traceable, verifiable documentation of data provenance, processing steps, and analytical decisions, consistent with GxP expectations, alongside robust safeguards for privacy and protection of sensitive data across the AI lifecycle. 
7. Model design and development practices: Best practices in software and system design should be used, supported by data fit-for-purpose and enabling interpretability, explainability, robustness, and generalizability. As the principles state, good development should promote “transparency, reliability, generalizability, and robustness for AI technologies contributing to patient safety.” 
8. Risk-based performance assessment: Evaluation should focus on the complete system, including human-AI interaction, using appropriate data, metrics, and validation methods aligned with the intended context of use, rather than algorithmic accuracy alone. 
9. Life cycle management: AI systems should be managed through ongoing, risk-based quality systems that support continuous monitoring, issue identification, and periodic re-evaluation, including addressing changes in performance such as data drift. 
10. Clear, essential information: Plain language should be used to communicate relevant information to users and, where appropriate, patients, including the system’s purpose, performance, limitations, underlying data, updates, and any interpretability or explainability features. 

Beyond guiding developers and applicants, the principles are intended to support wider international collaboration by highlighting opportunities for regulators, standards organizations, and other bodies to work together on research, education, harmonization, and the development of consensus standards that could inform future regulatory policy across jurisdictions. The initiative builds on ongoing EU–US regulatory cooperation following bilateral discussions in 2024 and aligns with EMA’s wider strategy on data, digitalization, and AI, including the European medicines agencies network strategy to 2028. Within the EU, further guidance is already in development, building on EMA’s 2024 AI reflection paper and reflecting evolving pharmaceutical and digital legislation. 

Commenting on the collaboration, European Commissioner for Health and Animal Welfare Olivér Várhelyi described the principles as: 

“A first step of a renewed EU–US cooperation in the field of novel medical technologies. The principles are a good showcase of how we can work together on the two sides of the Atlantic to preserve our leading role in the global innovation race, while ensuring the highest level of patient safety.” 

Reflecting on the publication, Peter Arlett characterized the principles as an enabling foundation rather than a final destination, stating: 

“This is a small but important step in our journey to turn AI promise into AI practice. These principles lay the foundation for detailed guidance, supporting the use of AI across the medicine lifecycle.” 

As the joint principles note, “as the use of AI in drug development evolves, so too must good practice and consensus standards.” Sustained international collaboration with public health stakeholders is therefore seen as essential to supporting responsible innovation and promoting global convergence in the use of AI across medicines development.