AI rules converging fast

Drug regulators in the United States and Europe are starting to use the same playbook for artificial intelligence in medicine: classify risk, prove the model is credible, and control it across its full life. (fda.gov) (ema.europa.eu) (digital-strategy.ec.europa.eu) The United States Food and Drug Administration issued draft guidance in January 2025 for drug and biological product submissions that use artificial intelligence to support a regulatory decision on safety, effectiveness, or quality. The agency said sponsors should assess model credibility based on the model’s role, the decision’s risk, and the consequences of error. (fda.gov 1) (fda.gov 2) The European Union’s Artificial Intelligence Act entered into force on August 1, 2024, with obligations phasing in over time, and the main rules for high-risk systems apply from August 2, 2026. Under Article 6, an artificial intelligence system can be high-risk if it is itself a regulated product, or a safety component of one, and that product needs third-party conformity assessment under European Union law. (commission.europa.eu) (ai-act-service-desk.ec.europa.eu 1) (ai-act-service-desk.ec.europa.eu 2) The convergence is showing up in joint documents, not just parallel speeches. In January 2026, the European Medicines Agency and the Food and Drug Administration published 10 principles for “good artificial intelligence practice” covering the medicines lifecycle from early research and clinical trials to manufacturing and post-marketing safety monitoring. (ema.europa.eu) (fda.gov) In plain terms, regulators are treating an artificial intelligence model less like a one-off spreadsheet and more like a manufacturing process that has to stay under control after launch. The joint January 2026 principles call for defined intended use, data quality, model design fit for purpose, performance monitoring, and documented governance over time. (ema.europa.eu) (fda.gov) That approach fits how drug companies are already using these tools. The Food and Drug Administration says artificial intelligence now appears across the drug product lifecycle, including nonclinical work, clinical development, postmarketing activities, and manufacturing, and the agency has seen a significant increase in submissions with artificial intelligence components in recent years. (fda.gov) The practical problem for companies is that one model can touch several legal buckets at once. A system used to analyze trial data, support manufacturing quality decisions, or monitor safety signals after approval may face Food and Drug Administration expectations on credibility and documentation while also falling under European Union rules for high-risk artificial intelligence, depending on how it is used. (fda.gov) (ai-act-service-desk.ec.europa.eu) (digital-strategy.ec.europa.eu) The European Medicines Agency has also been building out the implementation side. Summary notes from a February 2026 meeting of the Heads of Medicines Agencies and European Medicines Agency artificial intelligence group said the January 2026 principles were supported by a drafting group with 15 representatives from the European Medicines Regulatory Network and that work continues on a glossary of artificial intelligence terms. (ema.europa.eu) The rules are not identical, and the United States guidance is still draft while the European Union law is already on a fixed implementation clock. But the direction is now hard to miss: if an artificial intelligence tool helps make decisions about medicines, regulators increasingly want evidence that it is trustworthy before use and controls to keep it trustworthy after. (fda.gov) (ai-act-service-desk.ec.europa.eu)