AI, CGMP, and the Cost of Over-Reliance

On this episode of The Regulatory Mix, we examine the growing intersection of artificial intelligence, pharmaceutical manufacturing, and regulatory accountability. As AI moves deeper into Current Good Manufacturing Practice environments, the key question is not whether firms will use it, but whether they can do so without weakening quality oversight, critical thinking, and compliance discipline.

A recent FDA Warning Letter to Purolea Cosmetics Lab provides a clear example of what can go wrong when AI is used without meaningful human review. The firm relied on AI to generate product specifications, procedures, and production records, while basic CGMP responsibilities were not adequately understood or controlled. FDA specifically criticized the lack of Quality Unit oversight and the firm’s failure to recognize process validation requirements.

The episode also looks at FDA’s emerging risk-based framework for evaluating AI credibility. The level of assurance required depends on three key factors: the model’s context of use, how much influence it has on a decision, and the consequence if that decision is wrong. The greater the model’s role and the higher the patient or product risk, the stronger the expectation for validation and oversight.

Beyond compliance, the discussion turns to the effect of generative AI on human judgment. Research involving knowledge workers suggests that AI can shift cognitive effort away from direct problem-solving and toward verification and supervision. That can improve efficiency, but it also creates risks such as automation bias, reduced independent thinking, and skill atrophy when users become overly dependent on machine-generated outputs.

The takeaway is simple: AI can assist regulated work, but it cannot assume regulatory responsibility. Accountability remains with the firm, the Quality Unit, and the qualified individuals responsible for ensuring that every decision affecting product quality and patient safety is scientifically sound and compliant.

Reference Links: Purolea Cosmetics Lab Warning Letter: https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/purolea-cosmetics-lab-722591-04022026

FDA Draft Guidance on AI Model Credibility: https://www.fda.gov/media/184830/download

Research: AI and Critical Thinking https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf

Sponsored by Tailwind Pharma, LLC, a pharmaceutical quality and regulatory consulting firm specializing in FDA inspections, CGMP compliance, sterile manufacturing, data integrity, investigations, and regulatory strategy. Learn more at tailwindconsult.com.

Disclaimer: The views expressed in this episode are personal opinions for educational purposes only and do not constitute legal, regulatory, or medical advice.