Principles and rules on using artificial intelligence at EMA

The European Medicines Agency (EMA)'s staff follows clear principles and rules guiding the use of artificial intelligence (AI) tools and technology at work.
HumanVeterinaryAIData on medicines

Dutch is available via eTranslation, the European Commission's machine translation service.

Translate to Dutch | Important information about machine translation

EMA is using AI tools and technology to generate and modify content it is making available internally. Some of this content can also be adapted for external use. 

AI use at EMA is based on principles and rules that ensure human oversight and support compliance with the transparency requirements of the EU AI Act.

For more on EMA's work to use AI in medicine regulation, see:

Principles and rules on using Microsoft Copilot at EMA

EMA uses two versions of Microsoft's Copilot AI tool – Copilot chat and Copilot 365 – to help streamline the work and boost the productivity of its staff.

When using Copilot, EMA staff follow three principles:

  • Human oversight
  • Human-led decision making
  • Ethical use

A series of rules help EMA staff uphold the principles laid out above. These rules cover general as well as specific practice, such as for editorial and HR work. They include:

  • Knowing when to use and when not to use AI tools at work – for instance, using AI to summarise and analyse information, but not if this information is restricted, private or confidential
  • Ensuring EMA staff retain full responsiblity for the content they produce with AI
  • Using AI to inform and enable, but not substitute human decision-making when completing work tasks
  • Avoiding AI use that could cause harm to others – such as inferring someone's performance, emotional state, or personal characteristics
  • Not using AI for recruitment tasks such as CV screening and candidate ranking

For more information on how EMA staff use Copilot in their work, see: 

Guidance on creating and modifying content with AI

EMA is preparing guidance on use of AI to generate and modify content. 

This is in line with AI Act (Art 50) – which introduces transparency rules to foster a trustworthy information environment in which AI can be used widely​.

The guidance applies to:​

  • Content produced by or on behalf of EMA, including written content, speeches, presentations, images, audio and video​
  • both external and internal content

It sets out requirements and good practices for using AI in content production, including:​

  • Human review​
  • Disclosure​
  • Responsibility for the final content 

This guidance provides EMA staff with practical advice on how to produce, label and disseminate content with AI​. This enables EMA stakeholders to:

  • Recognise EMA's AI-generated or modified content, where it exists
  • Make informed judgements and interpretations

EMA expects to publish this guidance by end of 2026.

For more guidance on the use of AI in medicine regulation, see:

Page update history

An update log is available to show the date and summary of changes to this webpage. It does not include updates to linked documents or minor edits like typos or broken link fixes.

The tracking of updates begins in November 2026.

11 November 2026

Page first published

Share this page