From Locked to Learning Algorithms: Regulatory Governance of Adaptive AI/ML Models in Pharmaceutical Lifecycle Management

Artificial intelligence and machine learning (AI/ML) technologies are increasingly embedded across the pharmaceutical lifecycle, influencing activities ranging from early drug discovery to post-marketing pharmacovigilance. Regulatory frameworks governing pharmaceutical products and processes have historically been designed for static systems whose behavior remains unchanged following validation and approval. In contrast, adaptive AI/ML models are capable of continuous learning from new data, enabling dynamic performance improvement while simultaneously challenging established regulatory principles related to validation, reproducibility, transparency, and accountability. This paper examines the regulatory governance implications of transitioning from locked to learning algorithms within pharmaceutical lifecycle management. By synthesizing current regulatory perspectives from major jurisdictions and aligning them with principles of Good Regulatory Practice, this study identifies key governance gaps and proposes the need for lifecycle- based regulatory oversight. The analysis highlights how adaptive AI/ML systems necessitate a shift from static approval models toward continuous, risk-based governance frameworks that ensure patient safety while supporting technological innovation.