Ten Years of AI Medical Device Regulation: Policy Evolution and Industrial Impact Across the World's Five Major Markets

Introduction

Over the past decade, the integration of artificial intelligence (AI) into medical devices has been profoundly reshaping global models of diagnosis and treatment. From real-time sepsis prediction to automated retinopathy screening, AI-assisted diagnostic tools have continued to emerge. However, the adaptive nature and "black-box" characteristics of AI algorithms pose severe challenges to traditional medical device regulatory frameworks.

Between 2015 and 2025, major markets such as the United States, the European Union, China, Japan, and South Korea successively introduced targeted policies, attempting to strike a balance between innovation and patient safety. Yet there remain significant gaps in global regulatory standardization, and cross-border approval is complex and costly.

This article, based on a systematic review study published in *Frontiers in Medicine*, examines the evolution of global regulatory policies for AI medical devices (AIMD) over this decade and analyzes their impact on the medical technology industry.

Industry Background

AI medical devices have become a core growth area in digital health. According to industry observations, the application of machine learning-based medical devices in diagnosis, monitoring, and treatment planning is expanding rapidly. However, the problem of regulation lagging behind technological development is becoming increasingly prominent.

A key background factor is that deep learning models often lack transparent decision-making processes and may be continuously updated after deployment, breaking the traditional assumption that medical devices remain fixed once on the market. Therefore, regulatory agencies need entirely new approaches to assess safety, effectiveness, and algorithmic stability.

Against this backdrop, major markets began to establish their own regulatory pathways. These pathways not only affect corporate market-entry strategies but also shape capital flows and hospital procurement decisions.

Key Developments

United States: Progressive Regulation from Pre-Cert to PCCP

The U.S. FDA has long been at the forefront of AI medical device regulation. In 2017, the FDA released the *Digital Health Innovation Action Plan* and piloted the Software Precertification (Pre-Cert) program, attempting to assess developer capabilities at the organizational level. In 2019, the FDA proposed a regulatory framework based on the Total Product Life Cycle (TPLC), with the core being the Predetermined Change Control Plan (PCCP), which allows manufacturers to predefine algorithm update parameters before market approval, thereby enabling flexible iteration within the approved scope.

The *AI/ML Software Action Plan* released in 2021 further strengthened this direction. In 2023, the FDA published the draft guidance for PCCP marketing submissions, clarifying the permissible conditions for algorithm modifications. In December 2024, the final guidance was officially issued, providing a clear compliance pathway for the dynamic updating of AI medical devices.

European Union: The AI Act and Risk Classification

The EU has adopted a more comprehensive legislative approach. The *Artificial Intelligence Act*, passed in 2024, classifies AI medical devices as high-risk systems, requiring compliance with strict data governance, transparency, and human oversight requirements. While this raises the threshold for market access, it has also become one of the most stringent regulatory models worldwide.

China: NMPA Technical Review Guidelines ### China: NMPA's Technical Review Guidelines

In 2022, China's National Medical Products Administration (NMPA) issued the "Guiding Principles for Registration Review of Artificial Intelligence Medical Devices," providing a clear technical framework for the filing of AI medical devices. At the same time, China has also actively participated in international regulatory coordination to improve the approval efficiency of AI medical devices.

Japan and South Korea: Flexible Adaptive Regulation

Japan's PMDA has developed an "adaptive AI regulatory framework" that emphasizes flexible regulatory measures based on the risk level of algorithms. South Korea's MFDS is also revising its guidelines and aligning them with international standards to support domestic AI medical device companies going global.

Market Impact

These regulatory policies are reshaping the medical technology market.

First, for large medical device companies, mechanisms such as PCCP mean that products can be continuously iterated after market launch, extending product life cycles. For example, imaging AI companies can regularly optimize algorithms based on real-world data without having to reapply each time.

Second, for startups, regulatory transparency reduces uncertainty. The strict compliance requirements of the EU AI Act may increase costs for small businesses, but they also channel more funding toward companies with strong compliance capabilities.

Meanwhile, hospitals, as end users, are increasingly paying attention to the regulatory status of AI devices. A study shows that approximately 43% of FDA-approved AI medical devices lack clinical validation data, prompting purchasers to make clinical evidence a key decision-making factor.

On the capital side, regulatory clarity is seen as an important factor in attracting investment. The policy frameworks in the United States, the European Union, and China are relatively clear, helping to create a predictable financing environment.

Challenges and Risks

Although global regulatory frameworks are gradually being established, challenges remain prominent.

• Insufficient data diversity: Fewer than 30% of AI medical devices disclose the demographic diversity of their training datasets. This can lead to algorithmic bias, with particularly poor performance among minority ethnic groups.
• Lack of clinical validation: Only about 28% of FDA-approved or recognized AI devices have undergone prospective testing, and a large number of products rely solely on retrospective data or simulation validation.
• Difficulty in international coordination: Regulatory standards vary significantly across regions, causing cross-border approvals to require repeated submissions and localization adjustments, increasing costs and time to market.

These risks are not only related to patient safety but may also undermine public trust in AI healthcare, thereby affecting long-term market penetration.

Future Outlook

• Looking ahead over the next 3-5 years, global regulation of AI medical devices will show several distinct trends:- Mechanisms like PCCP will be more widely applied: The U.S. FDA has already approved multiple devices based on PCCP, and the EU and Japan may follow suit, making dynamic regulation the norm.
• Accelerated international coordination: Organizations such as the International Medical Device Regulators Forum (IMDRF) are promoting a unified AI regulatory framework to reduce compliance costs.
• The central role of real-world evidence (RWE): Regulators will rely more on real-world data to continuously evaluate the safety and performance of AI devices, requiring companies to build strong data collection and analysis capabilities.
• The rise of China and Southeast Asian markets: NMPA's guiding principles are influencing other Asian countries, and China's massive clinical data resources will become an important battleground for AI algorithm validation.

From a capital perspective, regulatory clarity will continue to drive investment, especially companies that can demonstrate real-world clinical evidence and algorithmic fairness will gain more support.