Adaptive AI Change Control and Revalidation: A Risk-Based Lifecycle Framework for Pharmaceutical GMP, Medical Devices, and AI-Enabled Quality Systems
Abstract: Artificial intelligence (AI) is increasingly being integrated into pharmaceutical manufacturing, quality management systems (QMS), laboratory operations, process monitoring, medical-device software, and regulatory compliance activities. Unlike conventional deterministic software, adaptive artificial intelligence and machine-learning (ML) systems may change their outputs as data distributions evolve, models are retrained, parameters are updated, or operational environments change. These characteristics challenge traditional change-control and validation practices, which commonly assume that an approved system remains functionally stable until a formally authorised modification is introduced. In regulated environments, insufficient control of adaptive behavior may compromise product quality, patient safety, data integrity, traceability, and continued regulatory compliance. Conversely, indiscriminate full revalidation following every minor modification may create unnecessary operational burdens without proportionate improvements in assurance. This paper proposes an integrated, risk-based framework for adaptive AI change control and revalidation, combining conventional pharmaceutical and medical-device quality principles with model lifecycle management, continuous performance monitoring, change-impact assessment, and predefined acceptance criteria. The framework comprises eight interconnected elements: AI system classification, change detection, risk-based impact assessment, change authorization, tiered revalidation, controlled deployment, post-deployment monitoring, and lifecycle review. It distinguishes between software changes, data changes, model changes, changes in operational context, and emergent performance degradation. A proposed three-tier revalidation model?limited verification, targeted revalidation, and comprehensive revalidation?is presented alongside illustrative decision criteria, a change-control workflow, and practical applications in pharmaceutical manufacturing and AI-enabled medical devices. The framework is mapped conceptually to relevant quality risk management guidance, computerised-system controls, AI management standards, and medical-device regulatory approaches. It is a proposed framework rather than a prospectively validated method; empirical evaluation is required to establish its effectiveness. The approach aims to improve the proportionality, traceability, and scientific justification of AI lifecycle decisions while maintaining human accountability and patient protection.
Keywords: Adaptive artificial intelligence; machine learning; change control; revalidation; Software as a Medical Device
How to Cite?: Dr. Pushpinder Paul, "Adaptive AI Change Control and Revalidation: A Risk-Based Lifecycle Framework for Pharmaceutical GMP, Medical Devices, and AI-Enabled Quality Systems", Volume 15 Issue 9, September 2026, International Journal of Science and Research (IJSR), Pages: 2041-2050, https://www.ijsr.net/getabstract.php?paperid=SR260921085503, DOI: https://dx.doi.org/10.21275/SR260921085503