Qwyn AI
Platform
Services
AI Regulatory Navigator
Free Tools
How It Works
Outcomes
Blog

AI Compliance NavigatorSources checked Nov 16, 2026

Can we use AI for this? Get the answer, the rules, and the path to inspection-ready.

Describe what you want to do, and where you sell. We'll tell you if today's rules allow it, what an inspector will ask, and what to prepare.

We want to usetoforsold in.

Others ask about:
Redesign the use

Not as the decision-maker. Use it to assist a qualified person.

Draft EU GMP Annex 22 says generative AI and LLMs shouldn't be used in critical GMP applications. Today's rules don't ban it, but validating an LLM as the decision-maker is hard to defend. Let it prepare the review. QA decides and records the decision.

See your path to inspection-ready
5
Rules in forceValidate to these today
2
Drafts and plansDesign for these now
7
Inspector questionsWith the evidence to answer each

Your path

Four stages to inspection-ready

Each stage shows what to do, what an inspector will ask, and the rules for where you sell. Answer one question per stage to build your plan.

List this use in your AI inventory with an owner and its GxP impact. Include AI features inside vendor systems, not only models you built.

What an inspector will ask

Where is AI used in your GMP processes?
Evidence to showAn inventory with owner, GxP impact, and status for each use.

The rules behind it

Is this use in your AI inventory, with an owner?
AI use inventoryList every AI use in your GMP processes. Each row gets a verdict, the next action, and the rules that apply, using the same logic as the hub.Maps to 21 CFR 211.68(b), Part 11, EU GMP Annex 11, Annex 22 draft §1
XLSXApproved by Qwyn AI

Rewrite the use so the AI prepares and a qualified person decides. Write that boundary into the intended use, name the reviewer's role, and rate the risk.

Under the Annex 22 draft, an LLM can't be the decision-maker in a critical GMP application. Keep the decision with a qualified person.

What an inspector will ask

What does it do, and what happens if it's wrong?
Evidence to showAn approved intended use and a risk rating.
Who makes the decision, and where is it recorded?
Evidence to showThe named reviewer role and the decision record.

The rules behind it

Is the intended use written and approved, with a risk rating?
Intended use and risk worksheetScope check for model type, an intended use an inspector will accept, a risk rating, and the human oversight decision.Maps to ICH Q9(R1), 21 CFR 211.192, 11.10(a), Annex 11 §1, Annex 22 draft §1, §3
PDFDOCX3 pagesApproved by Qwyn AI

Set acceptance criteria before you test, on data the model never saw in development. Then prove the controls: accuracy checks, audit trails, and access.

The Annex 22 draft's test approach assumes deterministic outputs. An LLM gives different answers to the same input, so validate the review process around it, not just the model.

What an inspector will ask

How do you know it's accurate enough to rely on?
Evidence to showPre-set acceptance criteria and a test report on independent data.
Can you show who changed it, and when?
Evidence to showAudit trail on data, model versions, prompts, and configuration.

The rules behind it

Were acceptance criteria set before testing, on independent data?
Validation evidence checklistEvery artifact an inspector may ask for, by stage, with the US and EU rule for each. Includes a generative AI add-on.Maps to 21 CFR 211.68, 211.100, 211.192, Part 11; Annex 11 §1–§13; Annex 22 draft §2–§10
PDFDOCX2 pagesApproved by Qwyn AI

Route every model, prompt, or vendor update through change control. Monitor performance and input drift, and keep records of human review decisions.

What an inspector will ask

What happens when the model or vendor updates it?
Evidence to showA change control record with an impact and retest decision.
How would you know if it got worse?
Evidence to showPerformance and drift monitoring with defined triggers.

The rules behind it

Do model and prompt updates go through change control?
AI vendor qualification questionnaire21 questions to send before you switch on an AI feature, with a generative AI section and a scoring key.Maps to Annex 11 §3, §9, §10; 21 CFR 11.10; Annex 22 draft
DOCX2 pagesApproved by Qwyn AI

Your plan

0 of 4stages ready
  • 1Inventory and scope
  • 2Intended use and risk
  • 3Validate
  • 4Operate and monitor

Includes your gaps, the rule behind each, and the templates to close them.

QA
Talk it through with a GMP quality expert30 minutes · no slides

Bring this use case. You'll leave with a written plan for every gap.

Regulatory horizon

What changed, and what's coming

In date order, for where you sell. Get an alert the day a draft becomes final.

In forceDraft or plannedContext, not a requirement
Jan 2011
EU GMP Annex 11 revised
The computerised systems annex in force today.
Dec 2018
FDA data integrity Q&A, final
Sets data integrity expectations for drug CGMP.
Sep 2024
EMA reflection paper on AI adopted
AI across the medicinal product lifecycle.
Jan 2025
FDA draft guidance on AI for regulatory decisions
Still a draft. Covers drugs and biologics.
Jul 2025
EU GMP Annex 22 published as a draft
First AI-specific GMP annex. Consultation closed Oct 2025.
Feb 2026
FDA CSA guidance reissued
Device production and QMS software. Reference only for drugs.
Jun 2026
MHRA sets expectations for AI-written inspection responses
Inspectorate blog. Accuracy, review, sign-off.
Jul 2026
CDER plans AI manufacturing guidance
Two items on the 2026 agenda. Not bound to publish.
Oct 2026
EMA publishes workshop input on Annex 22
Input to the revision. Not adopted.
Nov 2026
FDA launches a 'Super Intelligence' hub
A new landing page. No new requirement.
Today
Nov 17, 2026
Not announced
Annex 22 adopted
The final text replaces the draft. Retest decisions follow.
Not announced
CDER AI and ML quality guidance
FDA's first drug manufacturing AI guidance.
Not announced
CDER CSA supplement for AI
Software assurance for AI in drug manufacturing.

Not legal advice. Qwyn AI tracks what regulators publish and shows its stated status. Confirm obligations with your regulatory and quality teams before you act.

Reference

Build your planAnswer one question per stage

Get your plan as a PDF

Your path, every gap, and the rule behind each, plus all 7 templates.

  • Your plan as a PDF, with every gap and the rule behind it
  • All 7 templates: inventory, worksheet, checklist, and more
  • An alert when a rule in your stack changes
Qwyn AI

AI-powered deviation investigation and CAPA management for life sciences manufacturers.

Platform

  • Features
  • Capabilities
  • Integrations
  • Security

Free Tools

  • Pharma AI Literacy Center
  • AI Maturity Assessment

Resources

  • Outcomes
  • Blog
  • AI Regulatory Navigator
  • Trust

Company

  • About
  • Careers
  • Contact
  • Press
© 2026 Qwyn AI. All rights reserved.
Privacy PolicyTerms of ServiceTrust
ISO 27001:2022

Qwyn AI provides AI-powered deviation investigation and CAPA management software for pharmaceutical quality teams. Our platform automates hypothesis generation using industry-standard frameworks like 6M, root cause analysis, and audit-ready report creation for GxP-regulated environments. ISO 27001:2022 certified. Integrates with Veeva Vault Quality, TrackWise, MasterControl, LIMS, and ERP systems.