Annex 22 is a proposed new annex to the EU GMP guidelines (EudraLex Volume 4), drafted by the European Commission, EMA and PIC/S. It was released for public consultation in July 2025 alongside revised drafts of Annex 11 (computerised systems) and Chapter 4 (documentation). At the time of writing it is still a draft: check EudraLex Volume 4 for its current status before relying on its text.
The consultation draft focuses on AI models used in critical GMP applications — those that can affect patient safety, product quality or data integrity. It expects the regulated company to define the model’s intended use, test it on independent, representative data against pre-defined acceptance criteria, understand how it reaches its output, monitor its performance in use and keep qualified people responsible for the decisions it supports. The draft favoured static, deterministic models for critical uses; commentary on the revision suggests the final scope may change, so treat its details as provisional.
Why it matters in GMP
Annex 22 signals how inspectors will look at AI in GMP: as a computerised system function that must be fit for its intended use, validated, controlled and overseen by people. Existing expectations still apply — Annex 11, Part 11, data integrity and ALCOA+ — so an AI feature that cannot show who decided what, on which evidence, will be hard to defend.
- Define the intended use and the risk of every AI function before using it in GMP work.
- Test against independent data and documented acceptance criteria.
- Keep a qualified person accountable for each decision the AI supports.
How Swat iQua Suite handles it
Swat iQua Suite does not use AI in GMP decisions today: every GMP record is created by a named person and every decision is e-signed, under one audit trail. AI assistance is on our roadmap, designed to suggest and never decide — any suggestion will be reviewed and accepted by a person and recorded in the audit trail, in line with the human-oversight thinking in the Annex 22 draft.