As AI continues to revolutionise the medical device industry, regulations are struggling to keep pace. While AI holds immense promise for improving patient outcomes, ensuring safety, and enhancing clinical efficiency, the evolving landscape of AI technologies including predictive and generative AI presents new challenges for regulators, industry stakeholders, and clinicians alike.

One of the primary challenges in regulating AI in medical devices is the volume of fragmented standards. The US Food and Drug Administration (FDA) currently has over 200 individual standards related to AI in healthcare, yet does not have a universally accepted framework or set of best practices. This lack of cohesion complicates the regulatory process, leading to potential confusion for companies trying to navigate the regulatory landscape. Efforts are underway to develop more unified, international guidelines for AI in healthcare, but achieving a consensus across diverse global stakeholders remains a significant hurdle. A collective shift toward standardised, evidence-based practices is critical in ensuring both the safety and efficacy of AI technologies across borders.

In the regulatory landscape, predictive AI and generative AI must be treated differently due to their distinct functionalities. Predictive AI has been used in medical devices since 1995 and involves locked algorithms that are designed to provide forecasts or diagnoses based on historical data. These systems are often considered 'trusted' and undergo rigorous premarket testing to ensure their performance. In contrast, generative AI creates new content or solutions (such as images, text, or data models) and is still relatively new in healthcare applications. Its use in medical devices raises additional regulatory concerns due to its more dynamic, evolving nature. As generative AI technologies continue to develop, regulators are tasked with crafting new guidelines that strike a balance between innovation and patient protection.

A key area of focus for AI regulations is the post-market life cycle. Unlike traditional medical devices, which can be static once approved, AI-based devices may evolve over time through software updates and refinements. This means regulators must account for continuous post-market review and oversight. Predetermined change control plans have become a crucial part of the regulatory framework, especially as AI algorithms can be updated or tuned after deployment. The question of how to balance innovation with patient safety remains central. Regulators and companies must work collaboratively to ensure that modifications to AI systems - whether they involve local adjustments for specific patient groups or broader updates - do not compromise performance or safety.