Artificial Intelligence in Regulatory Affairs: Transforming Drug Development, Regulatory Decision-Making, and Global Compliance

Abstract

Background: Artificial intelligence (AI) and machine learning (ML) are increasingly being used throughout
the drug development lifecycle, prompting regulatory agencies to develop new frameworks for their use
in evidence generation, clinical trial execution, and post-market surveillance.
Objective: To critically appraise current AI/ML applications across drug development and regulatory
decision making, compare the regulatory approaches adopted by the US Food and Drug Administration
(FDA) and the European Medicines Agency (EMA) and to explore the challenges of global regulatory
harmonization and compliance.
Data sources: Literature and guidance documents were searched in PubMed/MEDLINE, Scopus- and
Web of Science-indexed journals, Nature Portfolio publications and primary regulatory guidance from
FDA and EMA, published predominantly between 2021 and 2026.
Review Methods: We used a comprehensive narrative review format to accommodate the broad scope of
the topic, from applied AI/ML literature to regulatory guidance documents and health-policy commentary.
Evidence assessed for source authority, considering agency positions, and highlighting convergence and
divergence; not chronological summary.
Key Findings: The FDA has transitioned from discussion papers to draft and final guidance to the use of
AI in the regulatory decision-making process, device software lifecycle management and predetermined
change control, based on its own experience with over 500 submissions of AI components from 2016-2023.
The EMA reflection paper sets out a human-centred, risk-based approach covering the whole lifecycle of a
medicinal product. AI applications have demonstrated credible retrospective performance improvements
in clinical trial recruitment, risk prediction, and pharmacovigilance, but prospective validation and
demonstrated impact on trial success or patient safety outcomes remain scarce. The regulators’ own use
of generative AI tools has led to concerns over reliability due to hallucinations, continuing to underscore
the need for oversight by humans.
Conclusion: AI/ML is revolutionizing regulatory affairs in drug development but the current landscape of
regulatory structures remains in an early guidance stage of maturity, and global harmonization, algorithmic
transparency and prospective outcome validation continue to be the major open challenges to translate
the technical promise of AI into reliable regulatory practice.