A new open-access review published in AI and Ethics discusses validation, ethics and lifecycle governance of AI and machine learning in GxP applications.
This is one of the most relevant recent publications for GMP and pharma quality teams because it directly addresses the gap between high-level AI principles and practical implementation in regulated environments.
The paper compares regulatory approaches from FDA, EMA and the EU AI Act, including expectations related to AI in GxP, the draft EU GMP Annex 22, FDA’s AI credibility framework, data integrity, validation, lifecycle management, human oversight and generative AI risks.
The authors propose an Integrated Validation, Ethics and Lifecycle framework, called IVEL. The framework is not presented as an official standard, but as a structured way to organize AI validation and governance activities across the lifecycle.
Why this matters for pharma and GMP:
- AI validation cannot be treated as a one-time software test.
- AI systems require defined intended use, risk assessment, data governance, performance qualification, lifecycle monitoring and change control.
- Generative AI creates specific risks, including hallucination, prompt sensitivity, factual inconsistency and unclear reproducibility.
- AI outputs used in GxP decisions must remain traceable to data, model version, system configuration and human review.
- The paper reinforces that AI systems should be governed within the pharmaceutical quality system when they influence regulated processes.
Practical takeaway:
Pharmaceutical companies should start converting AI governance principles into practical SOPs and validation templates. For each AI use case, companies should define the context of use, risk level, data sources, model versioning approach, human review process, performance monitoring plan and requalification triggers.
This paper is especially useful for teams preparing internal AI policies, AI validation approaches, Annex 22 readiness plans or AI risk-assessment templates.
Source:
html: https://link.springe … 7/s43681-026-01218-9