FDA 10 Guiding Principles of Good AI Practice in Drug Development

The U.S. FDA 10 Guiding Principles of Good AI Practice in Drug Development give pharmaceutical companies a clearer framework for using artificial intelligence across the drug lifecycle.

Developed jointly by the FDA and European Medicines Agency (EMA), the guiding principles shift attention from AI capability alone to how AI is governed, assessed, documented, and maintained.

Image

Why Do the FDA 10 Guiding Principles for Good AI Practice in Drug Development Matter?

AI is moving deeper into drug development, from nonclinical research and clinical trials to manufacturing and post-marketing activities.

The Food and Drug Administration (FDA) says its experience between 2016 and 2023 already included more than 500 submissions containing AI components.

The AI principles framework does not function as a detailed regulation. Instead, it establishes a common foundation for good AI practice in drug development, while supporting future standards, international collaboration, and regulatory thinking.

The agency understands that AI governance is a central part of overall drug development governance.

The 10 FDA Principles Place Governance Around AI

The 10 guiding principles for drug development cover the full lifecycle of an AI application:

1. Human-centric by design
AI should align with ethical and human-focused values. Human judgement remains central to responsible use.

2. Risk-based approach
Validation, oversight, and risk mitigation should match the AI application and its context of use.

3. Adherence to standards
AI should meet relevant legal, ethical, technical, scientific, cybersecurity, regulatory, and GxP requirements.

4. Clear context of use
Sponsors should define exactly why an AI system is being used, what role it performs, and where its scope ends.

5. Multidisciplinary expertise
AI development and use require expertise spanning the technology, scientific application, and relevant drug development processes.

6. Data governance and documentation
Data provenance, processing, analytical decisions, privacy, and traceability need to be documented and controlled.

7. Model design and development practices
Models should use fit-for-purpose data and appropriate development practices, with attention to reliability, robustness, generalisability, and explainability.

8. Risk-based performance assessment
Performance assessment should consider the complete system, including human-AI interaction, relevant data, testing, and appropriate performance measures.

9. Life cycle management
AI does not stop being a regulatory concern after deployment. Sponsors should monitor performance, identify issues, assess changes, and watch for data drift.

10. Clear, essential information
Users and stakeholders need understandable information about an AI system's context, performance, limitations, data, updates, and interpretability.

View interview How Should Pharma Companies Audit AI Vendors?

How Should Pharma Companies Audit AI Vendors?

“This is one of the riskiest industries there is because you're dealing with human lives, you're dealing with experimental protocols, and you're dealing with regulatory bodies where you might not get another shot at that clinical trial."

Image

Context of Use is the Strategic Control Point for Good AI Practice in Drug Development

The fourth FDA principle for good AI practice in drug discovery deserves particular attention because context of use connects AI capability with regulatory credibility.

The FDA’s separate 2025 draft guidance on AI supporting regulatory decision-making also uses context of use as the basis for assessing model credibility.

The draft guidance defines credibility around whether an AI model performs reliably for a specific question and application.

This has an important practical consequence. A model should not be judged simply on whether it performs well in general. Pharma needs to establish whether it is fit for the specific decision it supports.

That will become more important as AI in drug development now moves from exploratory research into evidence generation and regulated development.

The Big Picture for Good AI Practice in Drug Development

The FDA’s 10 guiding principles emphasise a disciplined model of AI adoption for good AI practice in drug development.

AI cannot be isolated technology projects. Governance must be built around:

• defined use cases and accountability
• controlled data and documented development
• proportionate validation and performance monitoring
• ongoing human oversight and lifecycle review

The opportunity for using AI in drug development is substantial, but regulatory credibility will increasingly determine where AI can create durable and maximal value.

At Pharmatica, we track the technologies, regulatory shifts, and research shaping pharmaceutical development. Our Insights connect innovation with the evidence, governance, and strategic decisions that determine whether new approaches deliver measurable value.

Pharmatica: Insight. Connection. Impact.