Artificial Intelligence (AI) has been the leading area of innovation over the past couple of years, including its potential regulatory applications. The 2025 AI Observatory report, published by the European Medicines Agency (EMA) in May 2026, provides an update on ongoing initiatives to integrate AI into regulatory processes at both the EU and international levels.
The report has been drafted by the Network Data Steering Group (NDSG) of the European Medicines Regulatory Network (EMRN) and covers both human and veterinary medicines. We provide a summary of the main report’s contents to better describe the current status of AI integration across the medicines lifecycle.
Policy and Guidance on AI
The first chapter of the report addresses the marked evolution of policy and guidance on AI for regulatory uses that characterised 2025 and early 2026. More specifically, last year saw the entry into force of several obligations under the EU AI Act, i.e., those on prohibited uses (Article 5) and AI literacy (Article 4), followed by obligations on general-purpose AI (Chapter V) and governance structures (Chapter VII).
A key regulatory high-level document, “Guiding Principles of Good AI Practice in Drug Development”, was also jointly published by the EMA and the US Food and Drug Administration in January 2026. The document identifies ten principles that should always guide the application of AI across the entire medicines lifecycle, from research and clinical trials to manufacturing and safety monitoring.
Digital twins of patients are emerging as a potential breakthrough innovation aimed at simplifying and accelerating drug development, avoiding unnecessary animal experimentation and using real-world data to model many pathophysiological processes and their interactions with administered drugs. EMA’s Quality Innovation Group (QIG) published its preliminary considerations on Pharmaceutical Process Models in 2024, including a risk-classification decision tree for digital twins and similar models. The public consultation phase closed in March 2024, while the final document is still pending release.
The European Medicines Regulatory Network (EMRN) also undertook preparatory actions to support the development of domain-specific guidance across the medicines lifecycle. The increasing implementation of AI also required targeted revisions to other pieces of EU legislation, namely GMP’s Chapter 4/Annex 11 on computerised systems and the new Annex 22 on artificial intelligence, which address digitalisation and AI in GMP. Public consultation on the relevant EudraLex Volume 4 chapters and annexes closed in October 2025.
A new coordinated AI guidance roadmap should also be drafted by the NDSG as part of its 2026–2028 workplan, with a focus on implementing the EU Biotech Act, which is currently under discussion by European co-legislators. This new guidance should also address the use of AI in clinical development and pharmacovigilance. EMA also contributed to the drafting of the CIOMS Working Group XIV report on AI in pharmacovigilance (link), published by the Council for International Organisations of Medical Sciences.
Applications of AI
Early regulator-industry meetings in 2025 were the preferred forum for discussing emerging applications and the acceptability of AI in medicines development. These included mainly Portfolio and Technology Meetings (PTM) and Innovation Task Force (ITF) meetings, and, to a lesser extent, QIG meetings and Qualification Advice (QA) and Scientific Advice (SA) procedures.
Examples of emerging applications included the use of generative AI to assist with drafting regulatory submissions and technical documentation, or to generate answers to queries; AI support for evidence generation, outcome prediction, and patient selection; predictive stability modelling and shelf-life testing; and models of pharmaceutical processes, including digital twins.
Within the EU Network, information from industry exchanges will support the EMRN in building knowledge of AI applications, anticipating trends, and better preparing for future regulatory needs.
AI has already become part of many regulatory processes. The Scientific Explorer application, for example, supports searches related to EMA scientific advice procedures for human and veterinary medicines and provides information on initial marketing authorisation applications for human medicines.
The AI@MPA toolbox, developed by the Swedish Medical Products Agency, is a web-based suite of six AI applications. In 2025, it added a generative AI tool for document processing and answering regulatory questions, and new tools for drug-package similarity search and a chemical embeddings map, enabling read-across of preclinical and clinical safety data to new small molecules.
The NDSG adopted its new “Network AI Tools framework and catalogue” in 2025, listing available tools that support the sharing and development of AI applications across the Network. It also identified 61 use cases, grouped under 4 main AI areas: drafting and summarisation of information, validation and quality assurance, knowledge mining and information retrieval, and other diverse use cases.
Regulatory collaboration on AI
The European Medicines Agencies Network Strategy to 2028 (EMANS to 2028) includes many actions aimed at enabling the transformative potential of AI. Initial initiatives primarily focus on training in generative AI.
The EU Agencies Network Working Group on AI (EUAN WG on AI) is another initiative aimed at supporting European regulators in implementing artificial intelligence through common approaches. The One Health Agencies (EMA, ECHA, ECDC, EFSA, EEA) also continued to host joint innovation workshops on a rotating basis.
At the international level, several regulatory forums on AI strategy were organised in 2025 by the International Coalition of Medicines Regulatory Authorities (ICRMA). EMA further collaborated with US regulators, for example, with its QIG directly involved with FDA’s CDER Emerging Technology Team and CBER. The European Medicines Agency also contributed to the ICH Reflection Paper on Advanced Pharmaceutical Manufacturing and participated as an observer in the European Blockchain Sandbox (Phase IV) on the generation of synthetic health data.
The HMA/EMA multi-stakeholder workshop on AI, organised in November 2025, was the main opportunity to exchange views with other stakeholders.
Research in regulatory sciences
Many projects aimed at further exploring AI’s potential in regulatory processes are part of the Horizon Europe, Horizon 2020 and EU4Health research programmes and are mapped in Annex 3 of the 2025 AI Observatory report.
EU-funded AI initiatives span the entire pharmaceutical lifecycle, from R&D and pre-clinical in silico predictions and disease modelling to drug repurposing and optimisation of clinical development.
In manufacturing, AI supports decentralised production and improved personalisation of many advanced treatments. Signal detection for pharmacovigilance and the use of real-world data are among the main applications in the post-marketing domain.
The report also identifies the gaps that still prevent the exploitation of AI’s full potential. These include the availability of regulatory-grade AI assessment tools, veterinary medicines-specific initiatives, and AI robustness, testing and continuous monitoring.
NDSG’s “Network AI Research Priorities” identifies seven domain priorities that should be reflected in the work of funding bodies and researchers. These include research integrity and intellectual property, accuracy and reliability of AI tools, data governance, confidentiality and consent, regulation and oversight, ethics, fairness and bias prevention, resources for adoption, and the expected impact on the workforce.
