Artificial intelligence (AI) continues to revolutionize healthcare, transforming how data is interpreted, diagnoses are made, and patient care is delivered. However, with these advancements come new regulatory challenges—particularly around ensuring safety, performance, and reliability over time. To address these issues, the U.S. Food and Drug Administration (FDA) has released a discussion paper through its Center for Devices and Radiological Health (CDRH), seeking public feedback on how to assess AI-enabled medical devices across their entire product lifecycle.
Why the Discussion Matters
The FDA’s latest call for input reflects a growing recognition that traditional evaluation models may not fully capture the unique risks of AI systems—especially those that continuously learn or adapt after deployment. The agency emphasized that while AI, including generative AI, holds tremendous potential to improve outcomes and accelerate innovation, it also introduces “new considerations” for maintaining ongoing safety and effectiveness.
These concerns center on performance drift, a phenomenon in which AI systems begin to produce inconsistent or biased outputs over time due to shifts in data inputs, clinical practices, or patient demographics. Unlike conventional medical devices that remain static once approved, AI-enabled technologies must be evaluated dynamically to ensure continued reliability in real-world use.
A Decade of AI Growth in Medical Devices
Since approving its first AI-driven medical device in 1995, FDA has authorized more than 1,000 such technologies—with submissions increasing tenfold since 2020. These tools range from diagnostic imaging software and predictive algorithms to clinical decision-support systems, all contributing to faster, more precise care. But as adoption grows, so does the need for consistent evaluation standards that adapt to evolving data and technology environments.
What FDA Is Asking For
The discussion paper invites input across six key areas:
- Performance Metrics: What indicators best capture safety, effectiveness, and reliability for AI-enabled devices?
- Real-World Evaluation: How can FDA and manufacturers develop frameworks to assess device performance outside of controlled testing environments?
- Data Quality & Postmarket Surveillance: What systems are needed to ensure the quality of real-world data sources used in ongoing evaluations?
- Monitoring Triggers & Response Protocols: When and how should developers respond to signs of performance drift or degradation?
- Human-AI Interaction: How do clinical workflows and user behavior affect device performance and patient outcomes?
- User Experience: What design and usability factors influence adoption, reliability, and safety over time?
FDA is accepting comments through December 1, 2025, and aims to use the feedback to guide future policy and regulatory frameworks for AI in healthcare.
The Path Ahead
The initiative builds on ongoing efforts by FDA’s Digital Health Advisory Committee to develop real-world performance monitoring strategies for AI devices. As healthcare organizations increasingly adopt adaptive algorithms, the agency’s goal is to balance innovation with accountability—ensuring that technological progress translates into tangible patient benefits without compromising safety.
EMMA International’s Perspective
At EMMA International, we help life sciences companies navigate the rapidly evolving landscape of AI regulation. From algorithm validation to postmarket performance monitoring, our team ensures that AI-enabled products meet both scientific and regulatory expectations throughout their lifecycle.
As FDA advances its framework for AI evaluation, organizations that integrate compliance into their design and monitoring strategies will be best positioned for long-term success. Our experts support clients in developing risk-based validation approaches, real-world evidence collection plans, and regulatory engagement strategies that align innovation with patient safety.
For more information on how EMMA International can assist, visit www.emmainternational.com or contact us at (248) 987-4497 or info@emmainternational.com.
Reference:
Eglovitch, J. S. (2025, October 6). FDA seeks input on evaluating AI-enabled medical devices. Regulatory Affairs Professionals Society.
For more information on how EMMA International can assist, visit www.emmainternational.com. Contact EMMA International at (248) 987-4497 or by email at info@emmainternational.com to learn more.
