DIA: Regulators tout international efforts to set AI policies
PHILADELPHIA – Officials from the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) said that increased international collaboration is influencing the development of policies about artificial intelligence during the Drug Information Association's (DIA) Global Annual Meeting last week.
Meanwhile, pharmaceutical industry members are creating a set of best practices to effectively implement these policies in real-world applications.
These views were presented at a panel discussion on the evolving AI regulatory landscape in drug development across the US and the EU.
“We are presenting at a pivotal moment following several months of increased engagement and activity across both EMA and FDA from multistakeholder collaborative workshops toward publication of joint AI principles, and the launch of coordinated workplans across Europe, but what is noticeable is that we are moving beyond from early explorations toward a phase of more structured implementations and alignment,” observed Junyang Wang, director of US global regulatory and scientific policy at EMD Serono, who moderated the session.
Wang further noted that there is a growing trend towards international convergence and more alignment on high-level principles, as well as tensions between embracing innovations while ensuring compliance.
Anindita Saha, associate director for strategic initiatives at FDA’s Digital Health Center of Excellence within the Center for Devices and Radiological Health (CDRH), emphasized that the agency’s AI approach has been driven by collaborative engagement. She noted that “Six years ago, we had AI steering committees and workshops and discussion papers, where we were able to get feedback on our proposal [on AI].”
The agency subsequently issued a draft guideline in January 2025. (RELATED: AI in drug development: FDA draft guidance addresses product lifecycle, risk, Regulatory Focus 7 January 2025)
Saha explained that the guidance addresses seven key questions to help define the main inquiry. These questions include: What specific problem are you trying to solve? Why are you using AI? What are you ultimately trying to achieve? Additionally, she noted that it’s important to consider the context in which the AI model will be used. It is also important to assess the level of risk associated with the AI model. “We strongly recommend that you reach out to us early in the process” to discuss the use of AI models in drug development.
She added that FDA is working to finalize the guidance. The agency has received over 1,400 comments so far and is currently working through them.
Saha noted some recent collaborative efforts FDA participated in, including the joint FDA-EMA document establishing guiding principles for good AI practices in drug development spanning the product lifecycle that was released early this year. (RELATED: EMA, FDA issue joint AI guiding principles for drug developers, Regulatory Focus 14 January 2026)
In addition, FDA collaborated with Health Canada and the United Kingdom’s MHRA on a paper regarding Good Machine Learning Practices (GMLP) for medical devices. This paper was adopted in January 2025 by the International Medical Device Regulators Forum (IMDRF).
Joaquim Berenguer Jornet, AI implementation lead for EMA, stated that a key focus of the EMA's work plan is stakeholder engagement and international collaborations. The work plan is structured around five pillars: strategy and governance, data analytics, AI governance policy, data interoperability, stakeholder engagement, and guidance and international activities. (RELATED: EU regulators update work plan for leveraging Ai and big data, Regulatory Focus 11 March 2026)
Another priority of the work is a focus on AI literacy, which he said “is very important to help us build our skills. To lay the foundation on how AI can be used and seen.”
He added that throughout 2025, EMA engaged with national competent authorities through multiple collaborative sessions, collecting AI use cases, and exploring solutions across the network.
Martin Heitmann, a consultant for organizations in the life sciences and a volunteer with the International Society for Pharmaceutical Engineering (ISPE), discussed ISPE’s recent GAMP Guide on artificial intelligence in different good practice (GxP) areas. He highlighted several key concepts from the guide, including production and process understanding, a life cycle approach integrated with quality management systems, scalable life cycle activities, and fit-for-purpose data. These concepts are based on EMA and FDA guidelines on AI.
The guide was developed by an international, industry-led team of over 20 academic and pharmaceutical industry experts.
How can we bring all of this together and operationalize it for pharmaceutical companies? How can we ensure it works consistently?” Heitmann asked.
“What does it take to deploy AI in GxP? It takes AI capabilities that are suitable, SOPs aiming for life cycle control, data covering the context of use, capable and scalable hardware, and people with sufficient AI literacy. AI is horizontal in nature and cannot flourish in silos. The ISPE GAMP guide bridges the concepts and disciplines,” he added. “We provide the how-to layer and provide good practice guides to put this all together.”
