Good manufacturing practice: Multistakeholder workshop on expert contributions to artificial intelligence guidance development (Annex 22)

EMA's Good Manufacturing Practice (GMP) / Good Distribution Practice (GDP) Inspectors Working Group is organising a two-day workshop to help shape a risk-based approach to the use of generative artificial intelligence (AI) in medicines manufacturing.
EventHumanCompliance and inspectionsRegulatory and procedural guidanceResearch and development

Date

, -
, -

Location

Online
European Medicines Agency, Amsterdam, the Netherlands

Event summary

EMA's two-day workshop aims to gather expert opinion and evidence to inform the EU guidance on the use of artificial intelligence (AI) in medicines manufacturing. This guidance is referred to as Annex 22.

The first day of the workshop (30 June 2026) is organised as an open session where experts present their opinions and evidence. 

The second day (1 July 2026) is organised as a closed session where the Annex 22 drafting group at EMA reviews expert contributions.

EMA expects the workshop to produce a report with expert input.

A stakeholder consultation that EMA carried out in 2025 on its draft Annex 22 guidance suggested support for potentially enabling the use of technologies such as generative AI (GenAI) or large language models (LLMs) in medicines manufacturing.

The draft Annex 22 had indicated that dynamic, adaptive and probabilistic models - such as GenAI or LLMs - should not be used in critical GMP applications.

EMA is still considering the implications of the stakeholder consultation results. As part of this process, EMA's Annex 22 drafting group, in collaboration with its Quality Innovation Group, is seeking expert input on possible control and mitigation measures such as guardrails.

These measure are part of a proposed risk-based approach for manufacturers interested in using AI technologies in GMP applications.

Workshop objectives

EMA aims to achieve the following objectives via this workshop:

  • Collect expert insights on responsible AI requirements, risk identification and mitigation approaches, and practical safeguards
  • Identify key principles such as data governance, model evaluation, transparency, accountability, and human oversight
  • Identify specific considerations, challenges, and real-world constraints of the pharmaceutical industry
  • Identify specific approaches to responsible use of AI in GMP applications

Stakeholders invited

EMA has invited AI experts active in AI research to contribute their wide range of expertise to the workshop:

  • AI experts nominated by industry associations
  • Academics
  • European Medicines Regulatory Network experts
  • Other international regulatory authorities

Questions for experts

EMA has provided experts with questions divided into six topics.

Select the expandable panel below to access the planned questions for each topic:

  • How could we accommodate adaptive and probabilistic models in Annex 22?
  • What would be the validation paradigm for adaptive models that should be in annex 22?
  • Please comment on the application of a quality risk management approach to the entire lifecycle of such AI system models that may be used in GMP high-risk areas where there may be a low detectability of deficiencies but a high impact on the patient, and taking into account ICH Q9R1 principles. In applying quality risk management, what should the regulated user take into account when selecting and using an adaptive or probabilistic model?

  • To what extent can available guardrail mechanisms reliably prevent, detect, or contain hallucinations, incorrect recommendations, or fabricated data when dynamic or adaptive models, probabilistic models, and generative AI / LLMs are used within GMP workflows?
  • When guardrails fail or detect uncertainty, what mechanisms are required to ensure timely escalation and prevention of GMP impact?

  • What level and form of human-in-the-loop oversight is still required when guardrails are implemented
  • Is this oversight sufficient to ensure accuracy, traceability, and accountability?

  • How can guardrails remain effective when the underlying AI model evolves (in instances such as updates, retraining and drift)?
  • What controls are necessary to ensure guardrail performance remains validated over time?
  • What type and level of evidence (such as validation data, stress testing results and failure analyses) would be needed to justify that guardrails are a reliable risk mitigation measure enabling generative AI use in GMP functions?
  • Can you demonstrate the effectiveness of such guardrails or similar measures?

  • Do current guardrail approaches sufficiently address GMP high-risk areas such as data integrity, process control, auditability, and traceability - especially where generative AI systems may influence or automate decision-making?
  • Where do experts believe the limits of risk-based mitigation lie?
  • Are there classes of critical decisions where no combination of guardrails and oversight would be sufficient?
  • Are there scenarios in which the level of guardrail effort required to mitigate generative AI risks becomes disproportionate or operationally impractical, thereby indicating that such systems should not be used in certain critical GMP functions?
  • What other elements of an overall control strategy should be considered for inclusion in the Annex 22 to allow the use of dynamic or adaptive models, probabilistic models, and generative AI / LLMs in GMP applications?

  • Is there a need to address prevention and protection of AI systems from tampering or unauthorised access in Annex 22 or is this already sufficiently addressed in Annex 11?
  • What would you propose in terms of appropriate wording?
  • What risks arise when guardrail infrastructure is outside the manufacturer’s direct control?
  • How can supplier qualification, change-control visibility, and independent audit capabilities be preserved in cloud-based AI supply chains?

EMA asks topic leads to present their responses during the workshop by highlighting the following:

  • Areas of alignment with other experts
  • Points of divergence with other experts
  • Rationale underlying differing perspectives
  • Relevant practical experience or implementation examples

Registration

Participation to this workshop is by invitation only.

Live broadcast

EMA plans to broadcast this workshop live. 

Details will follow in due course.

Agenda

EMA will make a draft event agenda available in due course.

Share this page