A review of more than 40 publications indicates that data quality, validation, traceability and governance will be key to integrating AI into regulated pharmaceutical environments

Artificial intelligence (AI) is gradually transforming the pharmaceutical industry. It is doing so at various levels: it contributes to the optimisation of processes, predictive maintenance, real-time quality control, as well as document automation and improved logistics performance.

Against this backdrop, a new study led by the Burkinabe researcher Jean-Marie Ouédragogo, a professor at Mohammed V University in Rabat, has presented a critical overview of the applications, limitations and regulatory frameworks associated with the integration of AI into regulated pharmaceutical environments. It has thus established that AI can only become a reliable tool for the pharmaceutical industry if it can be validated, audited and governed within a regulated environment.

The study is based on a corpus of over 40 recent academic publications, institutional reports and regulatory documents covering everything from pharmaceutical manufacturing to Good Manufacturing Practice (GMP), regulatory issues, the supply chain and governance. Published in the journal Annales Pharmaceutiques Françaises, it highlights that AI facilitates the emergence of more connected and adaptive production models within the context of the new pharmaceutical landscape. However, these benefits are contingent upon the quality and interoperability of devices, the validation of models, the explainability of systems, cybersecurity and the regulatory requirements of the various regulatory bodies.

A recurring issue is that regulatory frameworks are not keeping pace with technological developments, and the industry is struggling to find a way for both to progress at a more similar rate. However, as Ouédragogo points out, regulation should not be seen solely as a brake on innovation, but also as an indicator of the actual level of maturity of artificial intelligence systems.

In this regard, he asserts that a system may incorporate advanced interpretation techniques; however, if it does not meet validation requirements, as well as other applicable standards, its use in a regulated pharmaceutical environment faces significant limitations.

‘The development of AI with pharmaceutical credibility requires systems that are validatable, controllable and auditable’

The paper argues that the performance of AI depends on the quality of the data on which the systems are built and used. Key principles identified include data quality, interoperability, data governance, model validation, compliance with GMP requirements and specific governance for AI. In other words, it is not enough simply to incorporate an algorithm into a pharmaceutical process; it must be possible to demonstrate how it works, what data it uses and how it is monitored over time.

In this regard, the study attaches particular importance to environments subject to Good Manufacturing Practice (GMP). The review notes that AI systems used in these environments must be validated, traceable and auditable. This has a significant implication for the industry: AI must be integrated into existing quality systems and governance processes. Consequently, the more critical the role played by AI within manufacturing or quality control, the greater the importance of aspects such as validation, monitoring and documentation.

In a similar vein, the review argues that the adoption of AI in the pharmaceutical industry requires a framework capable of ensuring that systems are valid, governable and auditable, as well as compatible with the levels of trust required in the healthcare sector. The regulatory challenge lies in establishing how such AI can be demonstrated to be reliable, how it is monitored, and how it is kept under control throughout its lifecycle.

“AI represents a major driver of transformation in the pharmaceutical industry, but its value depends on its integration into robust industrial, organisational and regulatory frameworks. The development of AI with pharmaceutical credibility requires systems that are validatable, controllable, auditable and compatible with the trust requirements inherent in healthcare activities,” concludes Ouédragogo.