ICH GCP R3 is now in English & EU languages. Book Now
  • AI ClinicalAI ClinicalNewRight Arrow
  • Clinical PracticeClinical PracticeRight Arrow
  • PharmacovigilancePharmacovigilanceRight Arrow
  • ManufacturingManufacturingRight Arrow
  • Laboratory PracticeLaboratory PracticeRight Arrow
  • DistributionDistributionRight Arrow
  • Medical DevicesMedical DevicesRight Arrow

About

EU GMP Annex 22 AI, Data Integrity & Compliance in GxP Environments is a thirteen-module programme that establishes the first European GMP framework for AI/ML in regulated environments. The course explains why deterministic CSV is no longer enough, and walks learners through Annex 22's core disciplines: intended use, training data governance, independent validation, explainability proportionate to risk, meaningful human oversight, lifecycle monitoring, and inspection readiness. Annex 22 is taught alongside Annex 11, the EU AI Act, FDA's PCCP and CSA guidance, and ICH principles, so learners can build a single, defensible AI compliance programme.

EU GMP Annex 22 is recommended for QA and validation managers, IT and AI system owners, data scientists working in regulated environments, internal auditors, digital transformation leaders, and regulatory affairs professionals across pharmaceutical, medical-device, and biotech operations. On completion, learners can classify AI systems by risk and select proportionate validation effort, establish data governance across training, validation, and operational data, design and execute model validation with independent test data and inspection-grade documentation, implement explainability methods proportionate to model criticality, operate meaningful human-in-the-loop oversight, monitor for model drift across the lifecycle, and lead inspection-ready Annex 22 audits.

Course Syllabus

  1. What Is Artificial Intelligence in GxP Context?
  2. The Spectrum of AI in Pharma and MedTech
  3. Why Traditional CSV Falls Short for AI
  4. AI Lifecycle - End-to-End
  5. Why AI Skills Are Now Core QA Capability
  6. Why AI in GxP Is High-Stakes
  7. Operating Principles for Trustworthy AI
  8. Stakeholders and Responsibilities
  9. QA's Role in AI Governance
  10. Validation Team's Role in AI
  11. IT and Data Science's Role in AI Governance
  12. Senior Leadership's Role in AI Governance
  13. AI Vendor Governance Considerations
  14. Common Pitfalls in Early AI Adoption

  1. What Annex 22 Is and Why It Matters
  2. Origin and Drafting Context
  3. Annex 22 Diagram - Position in the Framework
  4. Annex 11 vs Annex 22 - Working Together
  5. Annex 22 - Foundational Principles
  6. Scope - What Annex 22 Covers
  7. Scope - What Annex 22 Excludes
  8. Annex 22 Scope - Static vs Dynamic AI
  9. Static AI - The Default Position
  10. Dynamic AI - Restricted Use
  11. Generative AI Under Annex 22
  12. Annex 22 Section Structure - At a Glance
  13. Common Misinterpretations of Annex 22

  1. Why Scope Matters First
  2. First Question - Is It AI Under Annex 22?
  3. Second Question - Does It Affect GMP?
  4. AI Risk-Based Scoping
  5. Static vs Dynamic AI
  6. Static vs Dynamic Classification
  7. Generative AI in GMP - Specific Considerations
  8. Generative AI Scope Considerations
  9. Scoping Decision Process
  10. AI System Inventory - The Foundational Tool
  11. Boundary Cases - When Scoping Is Hard
  12. Industry Examples of Scoping
  13. Inspector Perspective on Scope
  14. Industry Trends in AI Scope Boundaries
  15. Documenting Scope Decisions Defensibly

  1. Why Risk-Based Matters
  2. ICH Q9 Foundations Apply to AI
  3. AI-Specific Failure Modes for Risk Assessment
  4. Patient Impact - The Primary Risk Dimension
  5. Risk-Based Validation Effort
  6. Process Criticality Considerations
  7. Decision Reversibility
  8. Detection - Can the Failure Be Caught?
  9. Risk Acceptance Decisions
  10. Likelihood Assessment for AI
  11. Severity Considerations Beyond Patient Impact
  12. Inspector Perspective on AI Risk
  13. Risk Assessment in AI Procurement

  1. Why Intended Use Is Foundational
  2. What Intended Use Means Under Annex 22
  3. Components of an Intended Use Specification
  4. AI Risk-Based Validation Effort
  5. Intended Use - Out-of-Scope Conditions
  6. Annex 11 vs Annex 22
  7. Model Definition Specifications
  8. Linking Intended Use to Validation
  9. Intended Use Patterns - Worked Examples
  10. Worked Example - Visual Inspection AI Intended Use
  11. Communicating Intended Use to Operators
  12. Worked Example - Deviation Triage AI Intended Use
  13. Worked Example - Process Control AI Intended Use
  14. Common Intended Use Pitfalls
  15. Intended Use Approval Process
  16. Operational Discipline - Staying Within Intended Use
  17. Intended Use Versioning Strategy
  18. Inspector Perspective on Intended Use

  1. Why Data Governance Drives AI Quality
  2. Data Governance Scope Under Annex 22
  3. AI Data Pipeline Quality Gates
  4. ALCOA+ Applied to AI Data
  5. ALCOA+ Extensions for AI Data
  6. Training Data Lineage
  7. Training Data Quality Review
  8. Training Data Approval and Versioning
  9. Data Lineage Tooling and Practice
  10. Training Data Bias Considerations
  11. Test Data Independence
  12. Test Data Composition Requirements
  13. Operational Data Governance
  14. Data Source Validation
  15. Labelling and Annotation Governance
  16. Synthetic Data Considerations
  17. Data Privacy and Security Considerations
  18. Data Governance in Practice
  19. Operational Data Capture Architecture
  20. Inspector Perspective on Data Governance
  21. Data Governance Tooling Considerations

  1. Why Development Discipline Matters
  2. From Data Science to Engineering Discipline
  3. Reproducibility as a Foundational Requirement
  4. Version Control - Code, Data, and Models
  5. AI Data Pipeline Quality Gates
  6. MLOps Platform Considerations
  7. Algorithm Selection Methodology
  8. Hyperparameter Tuning Discipline
  9. Bias Prevention During Development
  10. Training Pipeline Architecture
  11. Reproducibility Tooling Examples
  12. Training Documentation Requirements
  13. Model Card Documentation
  14. Reproducibility Verification
  15. Development Practices in Detail
  16. Cross-Validation Methodology
  17. Hyperparameter Search Documentation
  18. Adversarial Testing During Development
  19. Calibration Verification
  20. Continuous Integration for ML
  21. Development Governance Records
  22. Inspector Perspective on Development

  1. Why AI Validation Is Different
  2. Conventional CSV vs AI Validation
  3. AI Validation Lifecycle
  4. AI Lifecycle - End-to-End
  5. Validation Plan Components
  6. Data Qualification - Training Data
  7. Data Qualification - Test Data
  8. Model Validation Metrics Quad
  9. Acceptance Criteria Definition
  10. Boundary Case Testing
  11. Operational Performance Qualification
  12. Validation Documentation
  13. Validation Report Structure
  14. Evidence Traceability
  15. Re-Validation Triggers
  16. Validation Failures and Lessons
  17. Inspector Perspective on Validation
  18. Validation Cost Considerations

  1. Why Explainability Matters
  2. Explainability Definitions
  3. Explainability Spectrum
  4. White-Box Models
  5. Grey-Box Models
  6. Black-Box Models
  7. SHAP - Shapley Additive Explanations
  8. LIME - Local Interpretable Model-Agnostic Explanations
  9. Explanation Audiences and Formats
  10. Surrogate Models
  11. Attention Visualisation
  12. Explainability Proportionate to Criticality
  13. Documenting Explanations
  14. Transparency Beyond Model Interpretability
  15. Practical Explainability Implementation
  16. Counterfactual Explanations
  17. Real Explainability Challenges
  18. Inspector Perspective on Explainability

  1. Why Human Oversight Matters
  2. Human Oversight Tiers
  3. Designing Meaningful Oversight
  4. Avoiding Automation Bias
  5. Oversight Design Patterns
  6. Reviewer Selection and Qualification
  7. Oversight Documentation Requirements
  8. Reviewer Training Programmes
  9. Override Authority and Procedures
  10. Oversight in Operation
  11. Engagement Metrics
  12. Reviewer Calibration
  13. Oversight Quality Audits
  14. Oversight Workflow Implementation
  15. Escalation Mechanisms
  16. Oversight Failures and Lessons
  17. Inspector Perspective on Oversight
  18. Module 10 Summary - Human Oversight
  19. Oversight Cost-Benefit Analysis
  20. Cross-Functional Oversight Roles

  1. Why Lifecycle Discipline Matters
  2. Lifecycle Phases Under Annex 22
  3. Continuous Performance Monitoring
  4. Three Types of Drift
  5. Data Drift Detection
  6. Concept Drift Detection
  7. Performance Drift Detection
  8. Drift Investigation and Response
  9. Drift Threshold Selection
  10. Out-of-Distribution Detection in Operation
  11. Change Control for AI Systems
  12. Retraining Strategy
  13. Configuration Management
  14. Lifecycle Integration with Quality System
  15. AI System Retirement
  16. Lifecycle Failures and Lessons
  17. Inspector Perspective on Lifecycle

  1. Why Daily Discipline Matters
  2. What Inspectors Look For
  3. Inspection Findings Severity
  4. AI Inventory as Foundation
  5. Documentation Hierarchy
  6. Validation Evidence Organisation
  7. Inspection Response Practice
  8. Inspection Response Workflow
  9. During-Inspection Conduct
  10. Common Inspection Questions
  11. Pre-Inspection Preparation Checklist
  12. Defending Specific Predictions
  13. Findings Response and CAPA
  14. Building Inspection-Ready Programmes
  15. Internal Audits of AI Programmes
  16. Mock External Inspections
  17. Cross-Functional Fluency Development
  18. Document Currency Disciplines
  19. Inspection Coordination
  20. Post-Inspection Improvement Cycle
  21. Lessons from External AI Inspections
  22. Inspection Documentation Hierarchy

  1. Why Case Studies Matter
  2. How to Use This Module
  3. Manufacturing AI Failures
  4. Laboratory AI Failures
  5. Supply Chain and Pharmacovigilance
  6. Cross-Cutting Lessons

Course Benefits

Benefits cpd_points icon
CPD Points

Gain Continuing Professional Development (CPD) Points, accredited by The Faculty of Pharmaceutical Medicine of the Royal College of Physicians of the United Kingdom. These can be used to count towards the distance learning element of any scheme that comes under the umbrella of The Academy of Medical Royal Colleges or any other scheme for which there is mutual recognition.

Benefits certification icon
Certification

Receive a personal certificate to show your subject knowledge on course completion.

Benefits affordable icon
Affordable

You get excellent value through our cost-effective prices. We can also offer you group discounts on larger purchases.

Benefits flexible icon
Flexibility

The course saves you time through the convenience of online availability. This lets you complete the interactive course at your own comfort.

Benefits up_to_date icon
Keep Up to Date

You will stay up to date with any changes to EU GMP Annex 22, the EU AI Act, FDA Predetermined Change Control Plans, and ICH AI principles as our training courses are constantly monitored, reviewed and updated.

Benefits industry_experts icon
Learn from Industry Experts

The course content has been developed by industry practitioners and subject-matter experts to ensure that learners can apply the principles directly in their daily work. Every module follows a consistent rhythm — concept, regulatory context, real-world failure, audit perspective, preventive controls, and a knowledge check.


Our Certified Customers

novartis
NHS
takeda
roche
dhl

Learner Rating & Reviews

4.7
★★★★★★
Average Rating
536 global ratings
★★★★★
87.0%
★★★★★
5.0%
★★★★★
3.0%
★★★★★
3.0%
★★★★★
2.0%
RC

Working with Whitehall training for the last two years of partnership has been a very successful experience – I have fast access to all the GCP course...

SM

I have finalised the demo for the ICH-GCP E6 R3 refresher course. Overall, I liked the content and the interface. I also want to thank Whitehall Train...