What the FDA and EMA’s 10 AI Principles Mean for Patient Support
FDA and EMA jointly outlined 10 principles for good AI practice across the drug product lifecycle. While not written specifically for patient support, they raise important considerations for pharma teams deciding where AI belongs in patient engagement, what safeguards it needs, and how to manage it over time.
1. Remain Human-Centric by Design
Begin by identifying where AI can strengthen the patient experience and where human involvement matters most. Technology can make engagement more relevant and responsive, while people remain central when judgment, empathy, or intervention is needed.
2. Take a Risk-Based Approach
Establish oversight based on how the technology is being used. Optimizing the timing of an educational message, for example, calls for a different level of scrutiny than using AI to support a higher-risk patient decision. Patient support teams should match safeguards to the risk level of each use case.
3. Build Within Pharma Standards
Apply the same standards that already govern patient support programs. Privacy, security, regulatory requirements, and applicable GxP practices should remain part of how AI is evaluated and implemented.
4. Define AI’s Role in Patient Support
Outline what AI needs to do before deciding how to use it. That could mean identifying patients who need additional engagement, choosing a more relevant channel, or recognizing when intervention is needed. A clear role and guardrails keep AI focused on a specific patient support need rather than simply adding automation.
5. Build With Multidisciplinary Expertise
Pair technical expertise with a deep understanding of the patient, treatment journey, and support program. This helps ensure AI is built around real patient needs and how support is actually delivered.
6. Know the Data Behind the Decision
Assess whether you have enough relevant, high-quality data for the intended use before applying AI. FDA and EMA also emphasize traceable data sources and protections for sensitive information. Teams should understand where their data comes from and how it will inform decisions.
7. Look Beyond Model Performance
Understand what’s driving an AI recommendation, where the technology has limits, and whether it performs consistently across the patients it serves. These factors matter when AI begins shaping how patients are engaged.
8. Evaluate the Full Support System
Rate AI as part of the broader support system, including nurse case managers, care teams, and caregivers. Look at what AI identifies, what happens next, when a person steps in, and whether the full experience works as intended.
9. Monitor Performance Over Time
Audit how AI performs as patient behavior, populations, and programs change. Review results regularly and adjust when performance begins to shift to ensure the technology continues to work as intended.
10. Make AI Transparent and Understandable
Communicate AI’s role, performance, limitations, and relevant updates in clear, accessible language. Patients and program users shouldn't need technical expertise to understand the information that matters to them.
What the FDA EMA AI Principles Mean for Patient Support
At Medisafe, we’re already using AI to help pharma teams respond to patient behavior with more relevant engagement. It can help identify when a patient may need support. It can also help determine what type of outreach makes sense and when a person should step in.
Source: U.S. Food and Drug Administration (FDA) and European Medicines Agency (EMA), Guiding Principles of Good AI Practice in Drug Development, January 2026.
