The Ten "Commandments" of AI in Drug Development

Guiding Principles of Good AI Practice in Drug Development, January 2026 (https://www.fda.gov/media/189581/download)

In January 2026, the U.S. Food and Drug Administration, together with international regulatory partners, published ten principles for Good AI Practice in Drug Development:

1. Human-centric by design

2. Risk-based approach

3. Adherence to standards

4. Clear context of use

5. Multidisciplinary expertise

6. Data governance and documentation

7. Model design and development practices

8. Risk-based performance assessment

9. Life cycle management

10. Clear, essential information

These principles are presented as a foundation for further work rather than a finalized implementation framework. They describe what regulators consider important when AI is used to generate evidence across the drug product life cycle, without defining how these expectations should be met in specific technical or organizational settings.

At this stage, the document is intentionally high-level. Practical interpretation and operationalization will likely evolve through continued dialogue between regulators, industry, standards bodies, and technology developers.

https://www.raps.org/news-and-articles/news-articles/2026/1/ema-fda-issue-joint-ai-guiding-principles-for-drug

Familiar Concepts in a New Context

Many of the principles may sound familiar to professionals working with established clinical trial systems such as EDC, CTMS, or eTMF platforms. Concepts like risk-based approaches, lifecycle management, data governance, documentation, and adherence to standards are already part of everyday regulatory practice.

What appears different is not the concepts themselves, but the context in which they are now being emphasized. When AI is introduced, familiar expectations are applied to technologies that may behave differently from traditional, rule-based systems. This naturally raises questions about interpretation rather than compliance.

An Open Question Worth Considering

One possible way to read the guidance is to view it as an invitation to reflect:

• Which of these principles are already well understood and operationalized in existing clinical systems?

• Where might AI introduce additional considerations that are less explicit in traditional software development?

• How might established practices evolve as systems move from deterministic behavior toward more adaptive or probabilistic approaches?

These are not questions with immediate or universal answers. They depend heavily on context, use case, system design, and regulatory interaction.

Early Guidance, Not Final Instruction

Importantly, the FDA document does not claim to resolve these questions. Instead, it sets a shared reference point for future discussion and alignment. The absence of technical detail should not be read as a gap, but as recognition that good practice in this area is still emerging and will require time, experimentation, and collaboration to mature.

For now, the principles serve as a common language, useful for orientation, internal discussion, and education.

Closing Note

As AI continues to enter regulated environments, documents like this are likely to be revisited, refined, and expanded. Understanding them as living guidance, rather than fixed rules, may be the most appropriate way to approach them at this stage.

For readers involved in clinical systems, software development, or regulatory oversight, the principles offer a structured way to think about AI, without yet demanding definitive answers.

Disclaimer: This post reflects an educational interpretation of publicly available regulatory guidance and does not constitute regulatory or legal advice.