Module 22
EU GMP Annex 22 and Artificial Intelligence in GMP Manufacturing
Regulatory Status Notice: The content of this module is based on the draft EU GMP Annex 22 “Artificial Intelligence” published for public consultation on 7 July 2025 by the European Commission and EMA alongside revised drafts of Annex 11 and Chapter 4. Annex 22 had not been finalised as at May 2026. This module will be updated as the final text is published and as the international AI regulatory landscape for GMP-regulated environments continues to develop. Subscribers will be notified when content is updated.
- VIP Course
Course Overview
This course is the Pharmy Academy VIP module on the new regulatory framework for Artificial Intelligence in GMP-regulated pharmaceutical manufacturing. It addresses the draft EU GMP Annex 22 published for consultation on 7 July 2025 by the European Commission and EMA alongside revised drafts of Annex 11 (Computerised Systems) and Chapter 4 (Documentation), and the wider regulatory landscape for AI in pharmaceutical manufacturing as it stands at the time of course access.
Annex 22 is the first comprehensive GMP framework dedicated to the use of Artificial Intelligence in the manufacture of medicinal products and active substances. It applies specifically to static, deterministic AI and machine learning models used in critical GMP applications where the model output directly affects patient safety, product quality or GMP data integrity. Models that adapt their performance during use, and probabilistic models whose outputs vary for identical inputs, are not permitted within the framework’s critical application scope. This boundary is itself a substantive regulatory position and one that affects how AI is conceived, validated and deployed in pharmaceutical operations.
The course addresses what Annex 22 requires: intended use definition, model selection rationale, training data management and provenance, performance specification including independent test datasets, validation evidence, lifecycle management including drift monitoring, explainability requirements proportionate to the application, human-in-the-loop oversight expectations and the change control framework that applies to AI models in critical GMP applications. It also addresses how Annex
22 interacts with the revised Annex 11 framework for computerised systems, where the boundary between conventional computerised system controls and AI-specific controls sits, and how organisations should govern AI deployments that span both.
As a VIP module, this course includes scheduled one-to-one mentoring sessions twice monthly throughout the subscription year. These sessions allow subscribers to bring their own organisation’s AI deployment questions, validation strategies, supplier evaluations or inspection responses to a senior expert for direct discussion. This is the highest-value component of the Pharmy Academy programme and is reserved for VIP upgrade subscribers within the 12-month tier.
The course is designed for senior practitioners deploying AI in GMP-regulated environments and for the QA, regulatory affairs, validation, IT and senior leadership professionals responsible for governing those deployments. It is forward-looking content for a regulatory area that continues to develop, and it is framed accordingly: the course tracks Annex 22 from draft through final adoption and beyond, with content updates incorporated as the framework matures.
Learning Outcomes
By the end of this course, learners will be able to:
- Explain the regulatory basis for AI in GMP-regulated pharmaceutical manufacturing under EU GMP Annex 22, the revised Annex 11 and Chapter 4 framework, and the wider international regulatory landscape including FDA, MHRA, PIC/S and PMDA positions.
- Apply the Annex 22 scope boundary, distinguishing AI models used in critical GMP applications from those used in non-critical applications, and distinguishing static deterministic models from adaptive and probabilistic models.
- Describe the Annex 22 expectations for intended use definition, model selection rationale, training data management, performance specification, independent test dataset requirements and validation evidence.
- Apply the Annex 22 lifecycle management expectations including drift monitoring, performance verification in production, retraining decisions where applicable and the change control framework that applies to AI models.
- Recognise the explainability expectations proportionate to the application criticality and the practical methods used to demonstrate model interpretability and prediction transparency.
- Apply the human-in-the-loop expectations and recognise where the framework requires human decision authority, human review or human oversight of automated AI outputs.
- Explain how Annex 22 interacts with the revised Annex 11 framework for computerised systems, where the boundary between conventional CSV controls and AI-specific controls sits, and how organisations should integrate AI governance into the wider computerised system management framework.
- Apply Annex 22 thinking to common AI deployment categories in pharmaceutical manufacturing, including process analytical technology, manufacturing process control, predictive quality analytics, visual inspection and packaging quality control, predictive maintenance and laboratory data review.
- Recognise the data integrity implications of AI in GMP environments, including data provenance for training datasets, audit trail expectations for AI-influenced decisions, traceability between model output and quality decision, and the explainability evidence required to defend an AI-supported decision.
- Identify situations requiring escalation, including unauthorised model changes, drift indicators outside specification, performance degradation in production, training data integrity concerns, supplier-driven AI changes and inspection findings affecting AI deployments.
- Apply Annex 22 expectations within the local Pharmaceutical Quality System and within commercial and regulatory engagement on AI deployments.
Course Content
Note on Regulatory Status: This course tracks the Annex 22 consultation draft published in July 2025. Annex 22 had not been finalised as at May 2026. Content will be updated when the final text is issued. References to Annex 22 requirements throughout this module reflect the draft consultation text current at the date shown at the top of this document.
Artificial Intelligence in pharmaceutical manufacturing has moved from concept to deployment faster than the regulatory framework that governs it. AI-supported process analytical technology, AI-driven visual inspection, AI-supported predictive quality analytics, AI-supported batch release decisions, AI-supported predictive maintenance and AI-supported laboratory data review are all in current use across the industry at varying stages of maturity. Until July 2025, the regulatory framework for these deployments rested on Annex 11, GAMP 5, FDA’s evolving CSA position, FDA’s 2023 discussion paper on AI in drug manufacturing, EMA’s 2023 reflection paper on AI in the lifecycle of medicinal products, and a wider patchwork of agency guidance. None of these was a comprehensive GMP framework dedicated to AI.
Annex 22 changes that landscape. As the first dedicated regulatory framework for AI in GMP manufacturing, it sets expectations that go materially beyond what Annex 11 alone provided. The framework’s restriction of critical GMP applications to static deterministic models, with adaptive and probabilistic models excluded from the critical scope, is itself a position that constrains how organisations can deploy AI. The expectations for training data provenance, independent test datasets, performance specification, drift monitoring and explainability define a validation framework that does not map directly to conventional CSV practice. The interaction with the revised Annex 11 and the broadened scope of computerised systems controls in the new framework adds further complexity.
Organisations deploying AI in GMP environments now need to demonstrate compliance with a framework that is genuinely new. This is not an extension of existing CSV practice. It is a distinct framework with its own expectations, its own evidence requirements and its own governance demands. The professionals responsible for AI deployments in pharmaceutical operations, and the QA, regulatory affairs and senior leadership professionals overseeing those deployments, need to develop framework-specific competence quickly.
The VIP positioning of this module reflects the strategic importance of getting AI deployment right in a regulated environment. The cost of an inspection finding against an AI deployment is significant. The cost of a Warning Letter or a non-compliance statement against an AI-supported critical quality decision could affect product supply, regulatory standing and commercial viability. The cost of a poorly conceived AI deployment that does not survive its first inspection is substantial. The cost of mentoring time with a senior expert who can work through specific organisational deployments, validation strategies and inspection responses is reliably lower than the cost of managing the alternatives.
This course is essential because it provides the depth, the practical application and the direct expert engagement that AI deployment in GMP-regulated environments now requires. It is forward-looking content for an evolving framework, and the VIP one-to-one mentoring component allows subscribers to engage with that framework as it applies to their specific organisational context.
The regulatory framework for AI in GMP manufacturing
The course opens by establishing the regulatory landscape. This includes the draft EU GMP Annex 22 in detail, the revised draft Annex 11 in the areas where it interfaces with AI applications, the revised Chapter 4 in its data governance implications, FDA’s 2023 discussion paper on Artificial Intelligence in Drug Manufacturing, EMA’s 2023 reflection paper on AI in the lifecycle of medicinal products and the wider international landscape including PIC/S consultation status, MHRA positioning and PMDA activity. The course will be updated as Annex 22 progresses from draft through final adoption.
Annex 22 scope and boundary
A dedicated section addresses the scope boundary that distinguishes critical from non-critical applications and that excludes adaptive and probabilistic models from critical scope. The course explains the practical implications of this boundary for AI model selection, deployment design and validation strategy. Learners will understand which AI applications fall inside the critical scope, which sit outside it and how the boundary affects governance.
Intended use, model selection and training data
The course addresses the Annex 22 expectations for intended use definition, model selection rationale and training data management. This includes the requirements for training data provenance, data quality assurance, dataset documentation, the relationship between training data characteristics and model performance, and the principles of training data integrity that the framework establishes.
Performance specification and validation
Performance specification and validation evidence under Annex 22 is addressed in operational depth. This includes the requirement for independent test datasets, the principles of performance characterisation, the validation evidence required to demonstrate fitness for intended use, the relationship between validation evidence and the criticality of the application and the documentation expectations across the validation lifecycle.
Lifecycle management and drift monitoring
The lifecycle management framework is covered including drift monitoring approaches, the principles of performance verification in production, retraining decisions where applicable to the deployment, the change control framework that applies to AI model changes and the relationship between model lifecycle and product lifecycle.
Explainability and human-in-the-loop
Explainability expectations are addressed in proportion to application criticality. The course explains the practical methods used to demonstrate model interpretability and prediction transparency, and the human-in-the-loop expectations that the framework establishes for AI-supported critical decisions. This includes where human decision authority is required, where human review is required and where human oversight of automated AI outputs is required.
The Annex 22 and Annex 11 interface
A dedicated section addresses the interface between Annex 22 and the revised Annex 11. The course explains where the boundary between conventional computerised system controls and AI-specific controls sits, where the two frameworks overlap, and how an organisation’s computerised system management framework should accommodate AI deployments without creating governance gaps or duplicate controls.
AI deployment categories in pharmaceutical manufacturing
The course addresses common AI deployment categories with the framework applied to each. This includes process analytical technology and real-time release approaches, manufacturing process control, predictive quality analytics, AI-supported visual inspection and packaging quality control, predictive maintenance and AI-supported laboratory data review. Each is addressed with a worked example of how Annex 22 expectations apply.
Data integrity in AI environments
The data integrity implications of AI are addressed in their own right. This includes training data provenance and integrity, audit trail expectations for AI-influenced decisions, traceability between model output and quality decision, the application of ALCOA++ principles to AI-supported processes and the explainability evidence required to defend an AI-supported decision to an inspector.
Inspection readiness for AI deployments
The final section addresses inspection readiness for AI deployments. The course explains how inspectors are approaching AI in GMP environments, what the early inspection findings look like and how to prepare an AI deployment for inspection scrutiny across validation evidence, lifecycle management, governance, training and supplier oversight.
VIP one-to-one mentoring component
VIP subscribers receive scheduled one-to-one mentoring sessions twice monthly throughout the subscription year. These are confidential expert sessions where subscribers can bring specific AI deployment questions, validation strategies, supplier evaluations, regulatory engagement plans or inspection findings to a senior practitioner for direct discussion. The sessions are organisation-specific and subscriber-driven. The cumulative mentoring time across the year is substantial and represents the highest-value component of the Pharmy Academy programme.
- Senior quality, regulatory affairs and validation professionals responsible for AI governance in GMP-regulated pharmaceutical manufacturing.
- IT, digital transformation and automation leads responsible for AI deployment design, validation strategy and lifecycle governance.
- QPs and senior batch certification professionals whose certification decisions are or will be supported by AI outputs.
- Site Quality Directors, Heads of Quality and VP Quality roles accountable for AI deployment compliance, risk acceptance and inspection readiness.
- Process analytical technology specialists, manufacturing science leads and MSAT professionals deploying AI in process control, quality analytics or batch release contexts.
- Laboratory leads deploying AI in chromatographic data review, OOS investigation support, characterisation analysis or other regulated laboratory contexts.
- Regulatory Affairs professionals managing AI-related regulatory submissions, post-approval changes affecting AI deployments and agency engagement on AI strategy.
- Senior leaders, executive teams and board-level governance functions whose strategic decisions on AI investment and deployment require informed regulatory understanding.
- Internal auditors, compliance professionals and inspection readiness leads working with AI-deployed operations.
- Consultants, advisors and contract validation professionals supporting AI deployment in pharmaceutical client environments.
You will gain the depth of understanding needed to lead, govern or contribute to AI deployment in GMP-regulated environments with genuine regulatory authority. You will be able to interpret Annex 22 alongside Annex 11 and the wider AI regulatory landscape, design AI validation strategies that meet current and emerging expectations, defend AI deployments to inspectors, sponsor organisations or corporate governance and engage with regulators on AI strategy from a position of informed credibility.
The VIP one-to-one mentoring component allows you to bring specific organisational questions to a senior expert throughout the year. This direct engagement is the most significant developmental opportunity in the Pharmy Academy programme and is reserved for VIP upgrade subscribers.
Organisations benefit from senior professionals who can govern AI deployments through the regulatory transition the industry is currently navigating. The cost of inadequate AI governance in a GMP environment is high and rising. The cost of a Warning Letter or a non-compliance statement against an AI-supported critical quality decision could affect commercial supply and regulatory standing. The cost of being unable to defend an AI deployment in inspection is one of the more avoidable strategic exposures in the current landscape.
The VIP mentoring component allows the organisation’s senior AI governance professionals to access expert input on specific deployments without the duration, cost or contractual complexity of engaging dedicated consultants. The cumulative value across a subscription year is substantial.
- Comprehensive expert video content covering the draft EU GMP Annex 22, the revised Annex 11 and Chapter 4, FDA and EMA AI guidance, and the wider international regulatory landscape.
- Real-world case studies from AI deployment in pharmaceutical manufacturing, laboratory operations, packaging and inspection environments.
- Practical examples of AI validation strategy design, drift monitoring implementation, explainability evidence construction and regulatory engagement on AI deployments.
- Cross-functional scenarios showing how QA, IT, validation, regulatory affairs and senior leadership integrate to govern AI in GMP environments.
- Exercises on Annex 22 scope assessment, AI deployment categorisation, validation evidence sufficiency and inspection finding response.
- Scheduled one-to-one mentoring sessions twice monthly with a senior expert, confidential to the subscriber, throughout the subscription year.
- Inspection-readiness scenarios focused on defending AI deployments across the framework.
- Multi-choice assessment examination.
- Certificate of completion upon passing the assessment.
- Ongoing content updates as Annex 22 progresses from draft through final adoption and as the international AI regulatory landscape develops.
Course Details
Instructor(s):
Paul Palmer & Aneta Jell
Level:
VIP
Duration:
3.5 Hours
Type:
Instructor led
Launch Yourself Into The Future.
Join Academy Pharmaceutical Excellence to gain industry-leading knowledge and skills through our comprehensive courses designed for aspiring professionals.