AI GxP validation for regulated life sciences
AI Software

What the Draft Annex 22 Requires From GxP AI, Section by Section

The draft EU GMP Annex 22 (July 2025, not in force) section by section: scope, intended use, acceptance criteria, test data, explainability, operation, and how to check a tool.

Updated Cybroscape Technologies5 min read

Key takeaway

The draft EU GMP Annex 22 (July 2025, not in force) section by section: scope, intended use, acceptance criteria, test data, explainability, operation, and how to check a tool.

The draft EU GMP Annex 22 on artificial intelligence was published for consultation in July 2025 and is not yet in force. It applies to AI models used in critical GMP applications and allows only static models with deterministic output there; generative AI and models that keep learning in use are kept out of critical uses. Where a model is in scope, the draft expects a detailed intended use, acceptance criteria set before testing, independent and representative test data, explainability and confidence where relevant, and change control, monitoring and human review in operation. Below is what each section asks for, what the draft does not say, and how to check a vendor against it.

What the draft covers, and what it leaves out

Annex 22 sits alongside Annex 11, which still governs the computerised system as a whole. Annex 22 adds expectations for the AI model inside it when that model is used in a critical GMP application, meaning one with a direct impact on patient safety, product quality or data integrity.

For those uses the draft limits itself to static models (frozen after training, not adapting from new data during use) with deterministic output, so the same input gives the same result. Dynamic models, generative AI and large language models are not to be used in critical applications. In non-critical uses they can be, with a qualified person responsible for checking what the model produces.

That line may move. EMA's workshop report, published on 2 October 2026, records that the drafting group has discussed allowing generative models under a documented, risk-based control strategy; our summary of the workshop report covers what changed and what did not.

Intended use

The draft expects the intended use to be described in detail and approved by the process subject-matter experts: what the model does, the input data it will see including its expected variation and limitations, and any subgroups of input that need separate attention. Everything later in the annex is tested against this description, so a vague intended use makes every later step weaker.

Acceptance criteria

Performance metrics and their acceptance criteria are set before testing, not after. The draft gives examples such as a confusion matrix, sensitivity, specificity, accuracy, precision and F1 score, and expects the model to perform at least as well as the process it replaces.

Test data and its independence

Test data must represent the full range of the intended use, be large enough to support the conclusions, and be correctly labelled, with the labelling itself checked. It must be independent of the data used to train and tune the model, and the draft expects controls that keep the test data separate from model development.

Test execution

Testing follows an approved plan, deviations are recorded and assessed, and the results are documented against the pre-set criteria. This is ordinary validation discipline applied to the model, and it should feel familiar to anyone who has run an OQ.

Explainability and confidence

Where it matters for the decision, the model should be able to show which features drove its output, for example through feature attribution or highlighting. The model should also record a confidence score with each result, and low-confidence results should be flagged rather than passed through as if they were certain.

Operation

Once in use, the model and its configuration are under change control, with retesting when either changes. Performance is monitored, and so is the input data, so that a shift away from what the model was tested on is noticed. Where a person reviews the model's output as a control, the draft expects that review to be recorded, and treats it as part of the system rather than an informal check.

What the draft does not say

Several claims about Annex 22 circulate that are not in the text. It does not use the term model card, though a model card is a sensible way to hold the intended use and test evidence. It does not require a documented fallback for when the AI is unavailable; that comes from Annex 11's business continuity expectations. It does not certify vendors or tools, and no software is "Annex 22 compliant" on its own, because the expectations attach to your use of the model.

It also does not apply to every AI feature in every system. A tool that drafts a document which a qualified person then reviews, edits and signs is a different case from a model that decides a batch disposition. Your own risk assessment decides which side of the critical line a use falls on.

How to check an AI tool against it

  • Which decisions does the AI make, and which does it only propose for a person to decide?
  • For any model in a critical use, is it static with deterministic output, and can the vendor show its intended use, acceptance criteria and test results?
  • Is the test data independent of the training data, and how is that controlled?
  • Does each output carry a confidence indication, and are low-confidence outputs flagged?
  • How are model and configuration changes controlled and retested, and how is performance monitored?
  • Is human review recorded in the system, with the reviewer, the decision and the time?

Our free Annex 22 readiness check turns the draft's sections into questions you can answer for one model.

How GxP Copilot is built against it

In Cybroscape GxP Copilot the AI drafts and quotes facts from your source documents; it does not decide. GxP impact, GAMP category and risk are computed deterministically from facts a person has confirmed, and every AI draft goes through Human-in-the-Loop reviews before it counts, with approval by electronic signature. Prompts, checks and model settings are fingerprinted, so a change fails the build until it is re-approved and the fixed test set is re-run. How the platform governs its own AI is described on the AI validation page.

Where to go next

Explore Cybroscape GxP Copilot for AI-native validation, Cybroscape TraceDraft for source-traceable clinical documentation, or book a demo to see either on your own data.

Part of our guide to GxP AI — the regulatory framework, validation expectations and platform landscape.

1Bounded intended useWhat the model does, and what it must not be used for
2A frozen evaluation setHeld-out data, version-controlled, agreed before testing
3A release gate that holdsAcceptance criteria set in advance, results recorded
4A person decides what mattersConsequential calls are made and signed by a named person
5Drift monitoring in productionInputs, outputs and overrides watched, with authority to suspend
6A fallback that actually runsThe process continues with the AI switched off
Annex 22 · consultation draft
The six parts of validating an AI system for GxP use. They map onto what the EU’s draft Annex 22 asks for. Annex 22 is still a consultation draft, not yet in force.
See it on your own systemBring a URS or a system. We run GxP Copilot against it live on the call.

Sources

Primary documents referred to on this page, from their official publishers.

annex 22 ai softwaregxp ai annex 22 requirementseu annex 22 software complianceai gmp annex 22 validation

Keep reading

All articles
AI Software5 GxP AI Software Buying Mistakes That Cost Quality Teams Six FiguresThe five most expensive mistakes pharma teams make when purchasing GxP AI software, and the evaluation steps that would have caught each one before the contract was signed. · 5 min readAI Software12 Ways AI Is Actually Being Used in Pharma Quality TodayWhere AI is used in pharma quality today: validation drafting, deviation triage, batch and lab data review, regulatory monitoring and clinical writing. What each does, where a person still decides, and what is not AI at all. · 6 min readAI SoftwareAI vs Manual Validation: An Honest Comparison for Quality LeadersA side-by-side comparison of AI-assisted and traditional manual validation: what AI actually does better, what it does not, and where the real time and cost savings come from. · 6 min readAI SoftwareAI GxP Validation Software: What It Does and How to Evaluate ItWhat AI does in GxP validation software today, what Part 11, CSA, the GAMP AI guide and draft Annex 22 say about it, how to spot AI-washing, how to validate the AI features and what to ask in a demo. · 8 min readAI SoftwareAutomated Validation in Life Sciences: What AI Can and Can't ReplaceAn honest assessment of validation automation: which activities AI handles well, which still require human judgment, the human-in-the-loop model, and the real ROI of automating validation work. · 8 min readAI SoftwareThe Real Cost of Manual Validation: Why Life Sciences Teams Switch to AIThe hidden costs of manual validation (time, rework, opportunity cost, and inspection risk) with benchmarks, an ROI calculation framework, and guidance on when to make the switch to AI. · 7 min readGAMP 5GAMP 5 Second Edition: what changed and why it mattersAgile, critical thinking, CSA alignment: decoded for teams who ship validated systems. · 6 min readAI TransformationGenAI adoption in pharma: from pilot to production without governance driftThe playbook that keeps use-cases moving through the intake without turning into shelfware. · 5 min read

Bring a system. We'll show you the package.

Book a demo, or ask a validation lead a quick question on WhatsApp.

  • We reply within one business day
  • We run your URS live on the call
  • You get a clear plan for your system
Chat on WhatsApp

Book a demo

Four fields. The right specialist replies within one business day.

Your details stay private. We never share them.