EMA Publishes Workshop Report on AI in GMP Environment

The European Medicines Agency (EMA) has published a report stemming from a multi-stakeholder workshop held on June 30, 2026, focusing on expert contributions to the development of guidelines for Artificial Intelligence. The report consolidates input on potential regulatory prerequisites for the use of adaptive, probabilistic, and generative AI systems in GMP-regulated applications.

The workshop's findings are intended to inform the revision of Annex 22 to the EU GMP guidelines. Initially, Annex 22 was not projected to cover generative AI and Large Language Models (LLMs) that adapt their performance. However, industry representatives argued during consultation that these models should be permitted to foster innovation. The workshop explored how a risk-based approach, coupled with a robust control strategy, could enable the use of generative AI and LLMs in GMP applications.

The workshop aimed to gather structured and experience-based industry contributions on governance, control, and risk mitigation for AI systems throughout their lifecycle. Particular emphasis was placed on applying the principles of Quality Risk Management as per ICH Q9. Industry interest groups collaboratively developed a response to a prior industry questionnaire, addressing six key themes including technical reliability, human oversight, validation, and cybersecurity.

According to the report, participants reached a broad consensus that the acceptance of an AI application should be assessed based on its documented intended use, the associated risks, and the effectiveness of controls applied throughout its lifecycle, rather than the underlying technology itself. Common requirements included robust quality risk management, lifecycle-wide validation and monitoring, and risk-proportionate human oversight. The EMA anticipates that the workshop's outcomes will aid in establishing a practical and scientifically sound regulatory framework that supports innovation while safeguarding product quality and patient safety.