FDA and EMA align AI principles

The U.S. Food and Drug Administration and the European Medicines Agency have issued a joint set of ten principles for using artificial intelligence in medicine development. (ema.europa.eu) (fda.gov) The agencies said the principles apply across the medicines lifecycle, from early research and clinical trials to manufacturing and post-market safety monitoring. (ema.europa.eu) (fda.gov) Artificial intelligence in this setting can mean software that spots patterns in trial data, predicts how a drug behaves, or helps monitor side effects after approval. Regulators are focused on whether those systems are reliable enough to support decisions about safety, effectiveness, and quality. (fda.gov) (ema.europa.eu) The ten principles start with design choices: human-centric development, a risk-based approach, and compliance with legal, technical, scientific, cybersecurity, and regulatory standards. They also call for data that is fit for use and systems built with established software engineering practices. (fda.gov) (ema.europa.eu) The document then moves to how models should be understood and controlled. It says developers should document intended use, make performance understandable to users and regulators, and manage changes over time so updated models do not drift outside their validated role. (fda.gov) (ema.europa.eu) The agencies also emphasize lifecycle governance: clear accountability, secure data handling, ongoing monitoring, and testing that matches the model’s context of use and level of risk. That means a low-risk internal tool and a model used to support regulatory evidence would not be judged the same way. (fda.gov 1) (fda.gov 2) This is not the first AI policy either side has published. The Food and Drug Administration released draft guidance in 2025 on AI used to support regulatory decision-making for drug and biological products, and said that work drew on more than 500 submissions with AI components received from 2016 through 2023. (fda.gov 1) (fda.gov 2) The European Medicines Agency adopted its reflection paper on artificial intelligence in the medicinal product lifecycle in September 2024, after first publishing a draft in July 2023. That paper set out the agency’s broader thinking on how artificial intelligence and machine learning should be evaluated in medicines regulation. (ema.europa.eu 1) (ema.europa.eu 2) What changed here is the alignment. Instead of separate regional signals, drugmakers now have a shared U.S.-Europe reference point for how regulators expect artificial intelligence systems to be designed, documented, updated, and watched after deployment. (ema.europa.eu) (fda.gov) The agencies described the principles as a foundation rather than a final rulebook, and said future work could include research, educational tools, harmonization, and consensus standards. For companies building AI into drug development, the immediate message is that performance claims alone will not be enough without traceable data, documented controls, and human oversight. (ema.europa.eu) (fda.gov)