Overview of Regulatory Alignment
The recent joint FDA–EMA guidance “Guiding Principles of Good AI Practice in Drug Development” has sparked conversation across the industry. Do you view it as a major shift or dismiss it as “just another guidance”? The truth usually sits between those extremes. This blog series will examine the signals, which we view not as a disruption but as regulatory convergence; this convergence alone is strategically significant.
Keeping the following themes in mind can help a company create, standardize, or optimize its own digital strategy by taking an AI-leveraged system and asking:
- What is the AI system intended to do?
- How could it affect product quality, safety, or regulatory decisions?
- How should risk drive the depth of validation and oversight?
These questions are designed to define the principles that this three part series will explore:
-
- The strategic significance of the FDA–EMA alignment (Part 1).
- AI governance and existing GMP frameworks, in hopes of revealing a less burdensome path forward (Part 2).
- A practical roadmap for operationalizing AI within a GMP aligned Quality System (Part 3).
Significance of the Joint Release
The joint release by the FDA and EMA of “Guiding Principles of Good AI Practice in Drug Development” has generated understandable interest across biopharma. Some see it as a disruptive inflection point while others view it as just another nonbinding document in an already crowded guidance landscape. The more accurate interpretation likely lies between those extremes: this is a strategic alignment signal, not regulatory upheaval.
Over the past year, regulators on both sides of the Atlantic have been steadily clarifying their thinking. The FDA’s January 2025 draft guidance on AI to support regulatory decision making and the EU’s ongoing refinement of the AI Act reflect a shared trajectory toward risk based, human centered oversight. The FDA–EMA guidance’s principles consolidate this thinking rather than reinvent it.
The ten principles—covering human oversight, risk based validation, data governance, transparency, and lifecycle management—correspond with the EU AI Act’s tiered risk framework. While the EU’s AI Act remains the only binding AI law today, the significance of this joint statement lies in the convergence of regulatory expectations. For global organizations, this convergence reduces uncertainty and signals how guidance is likely to evolve in the future.
The new guidance’s principles also reinforce the FDA’s credibility focused approach, grounded in “context of use.” Both frameworks ask the same foundational questions (What does the AI system do? How could it impact quality, safety, or regulatory decisions?). Regulators are not chartered to answer the “how”; they define the “what.” It remains the responsibility of a company (and its third parties) to determine how controls should be scaled to the defined risk. These are familiar regulatory positions, not new ones.
What’s changing is not the destination but the clarity of the roadmap. Regulators are increasingly explicit that AI must be governed like any other high impact technology—through proportional controls, transparency, and accountable human oversight.
Strategic Risk Leadership Dilemma
In order to assess whether a strategic blind spot exists in your organizations, ask your senior leadership team the following:
“Could we explain our AI governance, including risk controls, validation, and lifecycle oversight, confidently to any inspector today? (Or, for that matter, can we explain it to ourselves?)”
If not, Parts 2 and 3 of this series will help you to build this foundation.
In Part 2, we’ll explore whether companies should “create” or “integrate.” Depending on the current maturity of a company’s Quality Management System (QMS)/Pharmaceutical Quality System (PQS), establishing a clear starting point is critical. CGMPs already contain frameworks that can be leveraged, but how can they be optimized? The next blog in this series, Good AI Practice Meets Timeless CGMP, will discuss this.
