EMA’s AI Journey: The Rise of Artificial Intelligence in Medicines Regulation

In a world where artificial intelligence (AI) is moving rapidly from theory to practice, regulators are tasked with keeping pace—not only by writing high-level policies but by documenting what actually happens on the ground. Recognizing this, the European Medicines Agency (EMA) published, on 9th July 2025 & 8 May 2025, two complementary annexes capturing practical experience and real-world applications of AI and machine learning (ML) across the medicines lifecycle.

AI Observatory report – Practical Observations from EMA Procedures (8 May 2025)

Stakeholders acknowledge increased AI interest in automating data capture, protocol optimization, and predictive models for patient outcome and trial recruitment. There is a call for globally harmonized machine learning principles and regulatory frameworks that can evolve alongside technological advances. This annex complements the Observatory report by compiling EMA’s practical experience of handling AI-enabled submissions and regulatory interactions in 2024.

Key observations include:

• Growing number of AI-related questions: Sponsors increasingly seek scientific advice on AI models used in data analysis, endpoint detection, and predictive algorithms.

• Variation in AI maturity: Some AI tools are prototype-stage; others are embedded in pivotal studies or post-marketing surveillance.

• Methodological focus: Sponsors often ask about validation datasets, handling of missing data, generalisability across EU populations, and model updates over time.

• Regulatory responses: EMA applies risk-based principles, asking sponsors to justify AI use, explain decision logic, and demonstrate robustness.

These practical case notes illustrate the EMA’s approach: balancing innovation support with rigorous scientific evaluation. The EMA is also piloting and deploying internal tools to make its staff’s work easier.

Tool / System	Purpose	AI Type / Technology	Status
ChatGPT@EMA	Provide AI chat capabilities to staff	Generative AI	Under pilot
Speech-to-text	Transcribe and translate audio	Speech AI (Azure Cognitive Services)	Under pilot
Parallel Distribution	Compare EMA product documents and industry leaflets	Azure AI Document Intelligence	In production
Anonymization tool	Anonymise personal data on EMA website comments	NLP (Azure Cognitive Services)	In production
PDF Vendor Invoicing	Extract vendor invoice data and send to SAP	Azure AI Document Intelligence	In production
ARTE (Automatic Referral Template Editor)	Populate templates automatically from internal databases	Azure AI Document Intelligence	In production
Automated email distribution	Send invoices for operational approvals	Azure AI Document Intelligence	In production
CCIDAR	Identify confidential commercial info in public EMA docs	Azure AI Document Intelligence	In development

Health Data Lab Pilots:

• EurEKA: extracts ADR data from product info.

• MNEMOSiNE: uses AI to prioritize safety signal reviews.

• AERGIA: automates adverse reaction adjudication.

• ERATO: screens scientific literature for safety.

• OWLS: reduces duplicate case reports.

These tools reflect how AI is quietly but powerfully improving staff productivity, freeing experts to focus on more complex regulatory science.

Review of AI/ML applications in medicines lifecycle (9 July 2025)

This document “AI/ML applications in the medicines lifecycle,” provides a systematic map of where AI and ML are applied, from drug discovery to post-marketing safety.

Early drug development and discovery

AI supports target identification, drug design optimization, and prediction of drug–target interactions.Other uses include:

• Knowledge graphs to identify complex biological patterns.

• External control arms: AI combines historical and real-world data to create comparator arms in single-arm trials.

• Modelling fixed-dose combinations and simulating their potential benefit.

Manufacturing and inspections

To maintain product quality and reduce operational risk:

• AI supports automated visual inspection of finished products.

• Predictive maintenance tools minimize downtime.

• AI models help optimize batch processes, aiming to reduce waste.

Clinical development

Within clinical trials, AI helps:

• Identify eligible patient subgroups, using imaging and lab data.

• Predict long-term outcomes from biomarker signatures (multi-omics data).

• Increase statistical power by adjusting covariates through ML.

Companies also reported exploratory work to use AI to:

• Identify trial sites with better recruitment potential.

• Draft better protocols faster through natural language processing (NLP).

Post-authorisation safety monitoring

AI plays a role in pharmacovigilance by:

• Screening social media for early safety signals.

• Reducing duplicate reports in databases.

• Automating literature monitoring for case reports.

These tools can identify emerging safety issues faster, supporting timely public health action.

 For more detail click on the below LINKS

• Annex AI Observatory report

• Annex AI/ML applications in medicines lifecycle