GxP and GMP are not two competing standards. GxP is the umbrella term for every “good practice” discipline across the pharmaceutical product lifecycle, and GMP is the one that governs manufacturing, so all GMP is GxP but not all GxP is GMP. For an engineering team, the more useful question comes next: does the system you are building (an AI model, for example) fall inside GxP scope at all? The answer depends on whether it directly affects product quality, patient safety or data integrity.
GxP vs GMP: one umbrella, several disciplines underneath
GxP stands for “Good x Practice”, where the “x” is a placeholder for the discipline: manufacturing (GMP), distribution (GDP), laboratory work (GLP), clinical trials (GCP) and others. The MHRA’s data integrity guidance defines GXP as a collective term covering GDP, GCP, GLP, GMP and GPvP (pharmacovigilance). Some sources extend the list to storage (GSP) and documentation (GDocP), each covering a different stage of the product lifecycle.
GMP sits inside that family and focuses on manufacturing: facility design, equipment qualification, production controls, batch documentation, quality testing and product release. ComplianceQuest describes GMP as addressing cleanliness, sanitation, hygiene, validated processes, complaint handling, record keeping and personnel qualifications. For cleanroom control, see cleanroom compliance in biotech and pharma.
FDA guidance on quality systems ties manufacturers’ obligations to the current good manufacturing practice (CGMP) regulations in 21 CFR parts 210 and 211, and states that FDA’s inspection programme stays focused on compliance with those regulations. What those parts demand in detail is covered in what cGMP requires under 21 CFR 210/211. For what the “c” adds, see cGMP versus GMP and our explainer on what current good manufacturing practice means for pharma teams.
GxP and GMP side by side
| GxP | GMP | |
|---|---|---|
| What it is | Umbrella term for all good-practice disciplines in the product lifecycle | One discipline under that umbrella, focused on manufacturing |
| Lifecycle coverage | Research (GLP), clinical trials (GCP), manufacturing (GMP), distribution (GDP), storage (GSP), post-market monitoring (GVP) | Facility design, equipment qualification, production controls, batch records, testing, release |
| US legal basis | Regulations including 21 CFR Part 11 | CGMP in 21 CFR parts 210 and 211 |
| EU computerised-system rule | EudraLex Volume 4, Annex 11, cited among GxP regulations | EudraLex Volume 4, including Annex 11 and Annex 15 (qualification and validation) |
| Core aim | Quality, safety, traceability and data integrity across regulated processes | Products consistently produced and controlled to predefined quality standards |
The disciplines differ in detail, but they share a recognisable core. Eupry lists quality management, personnel qualification, facilities and equipment, documentation and records, production and process controls, and quality control as fundamentals across GxP. MasterControl compresses the same idea into “5 P’s”: people, procedures, products, premises and process.
How do I know if my system is GxP or non-GxP?
A system is GxP-regulated if it has a direct impact on product quality, patient safety or data integrity. If it does not, it is typically classified as non-GxP. The classification matters because it decides the validation requirements, documentation standards and audit exposure the system carries.
The test is applied to intended use. Microsoft’s GxP guidance for Azure customers puts the burden in the same place: the customer works out which GxP requirements apply to its computerised systems based on intended use, then demonstrates them through its own qualification and validation processes. So the same model architecture can be GxP in one deployment and non-GxP in another.
Sources do not all use “non-GxP” the same way. MasterControl uses it for software that is not compliant or was not designed with compliance in mind, such as a system not built for regulated industries. Eupry uses it for systems with no direct impact on quality, safety or data integrity. We find the impact-based definition the stronger one for scoping. A tool not designed for compliance can still end up in a GxP process, and then the gap is a compliance problem, not a scope exemption.
Effort is not uniform inside scope either. The MHRA notes that simple electronic systems with no configurable software and no data retention, such as pH meters, balances and thermometers, may only need calibration, while complex systems need validation for intended purpose, with effort rising with complexity and risk. Annex 11 says the same for GMP: risk management across the system lifecycle, taking account of patient safety, data integrity and product quality, should drive decisions on how far validation and data integrity controls go.
A scoping checklist for an AI system
- Write down the system’s intended use, since the GxP requirements that apply are determined by it.
- Ask whether that output directly affects product quality, patient safety or data integrity; if it affects none of them, the system is a candidate for non-GxP classification.
- Check whether the system generates or stores data that a GxP decision relies on, since software that generates or stores GxP-relevant data counts as a GxP system.
- If it is in scope, identify which discipline governs the activity (manufacturing, laboratory, clinical, distribution) so you know which GxP and which regulator apply.
- Size the validation effort to complexity, configurability and the opportunity for user intervention, rather than applying one level to everything.
- Base your standards and acceptance criteria on a risk assessment, since Annex 11 expects manufacturers to justify them that way.
The detail of what makes data and records GxP (ALCOA+, audit trails, the validated lifecycle) is covered in what makes data and systems GxP.
Check scope before assuming constraints
Not every system in a pharmaceutical organisation falls under GxP. That makes the first check on any proposed AI application whether it falls under regulatory requirements at all.
Which rules apply to a GxP computerised system?
Microsoft lists FDA 21 CFR Part 11 and EudraLex Volume 4, Annex 11 among GxP regulations. Annex 11 applies to all forms of computerised systems used as part of GMP-regulated activities, and it defines a computerised system as software and hardware components that together fulfil certain functionalities. Its core expectations are that the application is validated, the IT infrastructure is qualified, and replacing a manual operation causes no decrease in product quality, process control or quality assurance. It also expects periodic evaluation to confirm that systems remain in a valid state.
GAMP guidance does not define a prescriptive method or a standard. Instead it gives pragmatic guidance, approaches and tools, with a patient-centric, risk-based approach. The second edition keeps the first edition’s framework and its ICH Q9-aligned quality risk management, and adds an appendix on AI/ML covering compliance integration and fitness for use in a GxP environment. ISPE also says the second edition puts patient safety and product quality ahead of compliance and encourages critical thinking. ICH Q9 itself, in its R1 revision effective from 26 July 2023, gives principles and tools for quality risk management across development, manufacturing, distribution, inspection and review.
Enforcement follows geography: the FDA in the US, the EMA and national authorities in Europe, and regulators such as the MHRA (UK), PMDA (Japan) and NMPA (China) elsewhere. What an inspection looks like in practice is covered in our guide to pharmaceutical inspections and compliance essentials.
The EU rules are being revised. The revision covers Annex 11 and Chapter 4 on documentation, and adds a new Annex 22 on AI in medicines manufacturing. What Annex 22 asks of model selection, training and monitoring is covered in validating AI continuously under Annex 11 and Annex 22. For the EU AI Act, see the EU AI Act and GxP.
Can a system be “GxP certified”?
A “GxP certified” label on a system cannot be verified. When evaluating vendors, Eupry advises looking for evidence you can check: IQ/OQ/PQ validation documentation, conformance with Part 11 or Annex 11, and real certifications such as ISO 27001. A GxP-ready product helps, but how you configure, validate, use and maintain it decides whether your operations meet GxP requirements. The MHRA is blunter about vendor paperwork: it says vendor-supplied validation data on its own, without regard to system configuration and intended use, is not acceptable.
What an in-scope AI system has to produce
Approval, audit and regulated-workflow review all need artefacts that a non-engineer (an auditor, a quality owner) can read and sign against. TechnoLynx’s AI engineering for life sciences and healthcare work treats that evidence as an engineering output: evaluations, governance packs aligned to named rubrics such as GxP references, a model-risk inventory and lineage for production decisions. It is not legal advice, regulatory sign-off or certification. Our HIPAA / GxP workflow evidence pack shows a worked example of such an artefact, and Cassandra performance for GxP AI validation evidence stores covers the storage side.
The software-specific requirements are set out in what GxP compliance requires for AI software in pharma, and the quality and trial use cases in AI for pharma compliance. Two related applications are covered in AI for cleanroom compliance and nitrosamine risk control.
Once a system is in GMP scope, changes to an established process must be controlled and documented. Our practical guide covers how to go about changing processes inside GxP scope.
FAQ
What is the difference between GLP and GxP?
GLP (Good Laboratory Practice) is one discipline within GxP, covering laboratory work. GxP is the collective term that also includes GMP, GCP, GDP and pharmacovigilance practice.
What are the 5 P’s of GxP?
MasterControl sums up the principles GxPs share as people, procedures, products, premises and process. In turn, these cover trained staff, documented procedures with investigated deviations, documented specifications, controlled facilities, and defined processes with robust change control.