The use of Artificial Intelligence (AI) and Machine Learning (ML) in pharmaceutical manufacturing in India and globally is showing great potential to improve efficiency, product quality, and adherence to Good Manufacturing Practices (GMP). But when these technologies are applied in regulated environments, they bring several challenges—like how to validate their use, ensure data accuracy and integrity, manage risks properly, and meet strict regulatory rules. This article looks at the current rules and guidelines related to AI/ML use in GMP areas, points out the gaps that still exist, and suggests what should be done in the future to improve the policies. It also highlights important regulatory bodies such as the US FDA, European Medicines Agency (EMA), and the UK’s MHRA, along with real examples and case studies where AI/ML has been successfully used under strict regulatory supervision.
In today’s pharmaceutical manufacturing, Artificial Intelligence (AI) and Machine Learning (ML) are being used more frequently to improve batch production, predict machine maintenance needs, monitor product quality in real-time, and control processes more efficiently. As the use of AI/ML grows in this field, regulatory authorities have started creating guidelines to maintain a balance between encouraging innovation and ensuring patient safety and high product quality. However, since AI/ML systems often behave unpredictably and their risks are still being studied, regulatory bodies remain cautious and are moving forward carefully.
“The future of AI is not about replacing humans, it’s about augmenting human capabilities.” — Sundar Pichai, CEO of Google
The US FDA is actively working with the pharmaceutical industry to use AI and ML in manufacturing. It started the Emerging Technology Program (ETP) to study and guide such new technologies. In 2021, it also launched a special program called FRAME to focus on AI/ML in pharma. By June 2025, the FDA plans to use AI in all its departments. In early 2025, they successfully completed their first pilot review using AI, showing a big step forward in using AI in regulatory work.
In 2021, the EMA released a paper discussing how AI can be used in pharma manufacturing and the rules around it, focusing on maintaining GMP standards and data accuracy. With the 2024 EU AI Act, any AI used in pharma will now be considered high-risk, needing strict checks and human supervision. EMA has also prepared an AI action plan until 2028 to manage AI’s future role in medicine manufacturing.
The UK’s MHRA started a program called AI Airlock to safely test AI in healthcare settings. It also supports AI systems for quality control in pharma, but they stress the need for proper validation and managing changes carefully. Through its Innovation Passport scheme, the MHRA gives faster regulatory support to new AI-based pharma technologies.
The ICH gives global guidelines that support quality control and risk management in pharma, which fit well with AI/ML use. Its Q9 guideline promotes advanced tools like AI for managing risks. Also, the Q13 guideline supports AI in continuous manufacturing by setting clear rules for process monitoring and control that work well with AI systems.
Main Regulatory Challenges
- Validation and Checking
One big challenge in using AI/ML in pharma is checking (validating) the system properly, because these models keep learning and changing. Unlike old systems, AI does not stay the same. So, rules now ask companies to make proper plans for how and when the AI system will be updated. New ideas like “dynamic validation” and “predetermined change control protocol (PCCP)” help in keeping track of the AI system’s performance over time. - Data Integrity (Honesty and Accuracy of Data)
AI systems must follow ALCOA+ rules, which make sure the data is correct, complete, and can be checked later. Some AI works like a “black box,” which means we can’t easily understand how it made a decision. This makes it hard to track and audit data. That’s why “explainable AI” methods are being used so the AI can show how it reached a decision in a simple and clear way. - Understanding and Clarity (Explainability and Transparency)
Regulators want AI systems to be clear and easy to understand, especially when they affect product quality or safety. Tools like SHAP and LIME are helping to make AI decisions more visible and understandable. Companies are advised to build AI models that are easy to explain from the beginning, so people can trust and check their results. - Managing Changes in AI
Since AI models keep learning and changing, there must be a system to manage these changes properly. Authorities expect companies to make a full plan for the AI’s life—from development to regular updates, checking, and control. A method called “progressive validation” helps in updating the model while following all rules without creating problems. - Ethical and Legal Issues
AI systems can sometimes be unfair or biased. For example, if the data used to train AI is not balanced, it can give wrong or unfair results. Regulators now want companies to check for bias and make sure AI works fairly for all. Companies must take responsibility for the results their AI models give and make sure they are honest and fair.
Easy Ways to Follow Rules While Using AI in Pharma
- Use Risk-Based Thinking
Regulators suggest using AI based on how risky the work is. If AI is used in something very important like medicine quality or patient safety, then it should be checked very strictly. But if AI is used in small jobs like making schedules, then it doesn’t need too much checking. - Keep Full Records
All details about the AI model—how it was made, what data was used to train it, how it performs, and what changes are made—must be written and saved properly. This is very important for audits and approvals by the regulators. - Human Checking Is Still Important
Even if AI is doing most of the work, people must still check what it is doing. Experts must be there to see AI’s results, understand them, and stop or change things if needed. AI should help humans, not replace them. - Set Up AI Monitoring Teams
Pharma companies should make special teams to look after AI use. This team should include people from quality, data, and compliance departments. Their job is to watch AI’s performance, control risks, and ensure everything is done as per rules.
Real Example of AI in Pharma
Janssen Pharmaceuticals showed how AI can be used properly. In 2016, they got approval from the FDA to use AI in making their HIV medicine Prezista. They moved from old batch production to continuous manufacturing. This change, with help of AI, made the process faster and better, reducing the time for testing and release, and allowed live monitoring.
What Should Happen in the Future
- Same Rules Everywhere
Right now, different countries have different rules. It is better if all countries follow similar rules so global pharma companies can follow one system. - Clear Rules for AI
Regulators should make new and simple guidelines only for AI. These should explain how to handle AI’s changing nature and continuous learning. - Train Regulators in AI
Officers and inspectors should be trained in AI so they understand how to check and approve AI models properly. - Build Trust with Openness
Companies should be clear about how their AI systems work. This builds trust with regulators and the public.
Using AI and machine learning in pharma manufacturing can bring big benefits like better quality, safety, and faster work. But there are also big challenges like how to check AI systems and keep data safe. The current rules are a good start, but we need to keep improving them. If the industry and regulators work together, AI can truly make medicine better and safer for everyone.
