The Impact of the European Commission's AI Regulation on Medical Devices

# Introduction

Artificial intelligence (AI), a transformative force in the modern era, has seamlessly woven itself into the fabric of society over the past few years. Its applications extend far beyond the realms of cutting-edge technology, reaching into everyday life, shaping work environments, and even adding layers of complexity to entertainment experiences. As the influence of AI continues to expand, so does the need for a comprehensive regulatory framework to govern its diverse applications. Recognizing the multifaceted impact of AI, the European Commission took a significant step forward by introducing a draft regulation in April 2021. This regulation, titled "Regulation laying down harmonizing rules on artificial intelligence (AI Act) and amending certain Union acts," signals the Commission's commitment to establishing uniform guidelines for AI across the European Union.

The decision to propose a regulation, rather than a directive, underscores the Commission's recognition of the urgent need for standardized regulations that can adapt to the rapidly evolving landscape of AI technologies. This article delves into the intricacies of the proposed regulation, specifically examining its implications for stakeholders in the field of medical devices and in vitro diagnostic medical devices. As AI becomes increasingly intertwined with healthcare technologies, understanding the potential changes brought forth by this regulatory initiative is essential for manufacturers, distributors, and importers operating in this dynamic and vital sector. In exploring the contours of the draft, we aim to shed light on how it could reshape the landscape of AI integration in medical devices and the ensuing challenges and opportunities for the industry.

# Defining Artificial Intelligence

Within the draft regulation, a pivotal aspect is the nuanced definition it ascribes to artificial intelligence (AI), marking a significant departure from previous attempts to encapsulate the expansive field. Art. 3 No. 1 of the draft introduces a precise legal definition, characterizing an "artificial intelligence system" as software developed using techniques enumerated in Annex I. This inclusive definition is deliberately broad, reflecting an awareness that legislation often lags behind the swift evolution of technology.

The deliberate broadness is an acknowledgment of the challenges posed by the rapid emergence of new technologies, ensuring that the regulatory framework remains adaptable to unforeseen advancements. However, the draft imbues this definition with specificity by referencing techniques and approaches outlined in Annex I. Among these are machine learning, deep learning, logic and knowledge-based concepts, and scientific methods. While this specificity aims to provide clarity, it also raises concerns about the potential breadth of application, potentially encompassing conventional programming methods with minimal risk.

The expansive nature of this definition, when coupled with the comprehensive list of techniques in Annex I, opens the door to a myriad of possibilities. It introduces a regulatory approach that seeks to be all-encompassing, but this breadth may inadvertently raise questions about the appropriate application of rules to diverse software types. Specifically, there is a concern that the all-encompassing nature of the definition may subject even low-risk conventional programming methods under Art. 3 No. 1, creating an environment of legal ambiguity and uncertainty for the developers of medical software.

It’s safe to be said, the draft's attempt to grapple with the dynamism of AI technologies is both commendable and challenging. Striking a balance between providing a forward-looking regulatory framework and ensuring specificity to avoid unintended consequences becomes a delicate task. As the regulation progresses, stakeholders will need to closely monitor how the definition evolves and whether the inclusion of various techniques strikes an appropriate equilibrium between innovation and regulatory clarity.

# AI Regulation and Medical Devices

Delving deeper into the regulatory landscape outlined in the draft, it becomes evident that medical devices are poised to undergo a paradigm shift in their compliance requirements. Art. 6 (1) in conjunction with Annex II categorizes certain medical devices as "high-risk AI systems." This classification is particularly noteworthy as it aligns with the existing regulations governing medical devices, such as Regulation (EU) 2017/745 for medical devices (MDR) and Regulation (EU) 2017/746 for in vitro diagnostic medical devices (IVDR).

The deliberate inclusion of medical devices within the ambit of the draft reflects a heightened awareness of the intricate intersection between artificial intelligence and healthcare technologies. Medical devices, especially those employing AI or incorporating AI components, are now subject to specialized scrutiny, acknowledging the unique risks and challenges they present. By listing MDR and IVDR in Annex II, the draft acknowledges the need for harmonization with existing legislation, streamlining the regulatory landscape for stakeholders operating in the medical device sector.

This strategic alignment with established regulations suggests a cohesive approach, aiming to build on existing frameworks rather than creating isolated silos for AI-specific compliance. The integration of AI into medical devices, a trend that promises enhanced diagnostic capabilities and improved patient outcomes, is met with a corresponding elevation in regulatory expectations.

As a consequence, the classification of certain medical devices as "high-risk AI systems" triggers a cascade of stringent requirements. These requirements span a spectrum of elements crucial to ensuring the responsible development, deployment, and ongoing management of AI-integrated medical devices. From robust risk management systems and thorough testing procedures to transparent documentation and human oversight, the draft places a multifaceted burden on manufacturers, distributors, and other stakeholders involved in the medical device ecosystem.

While this approach serves the overarching goal of safeguarding public health and ensuring the reliability of AI-integrated medical devices, it also poses a considerable challenge. The intricacies of conforming to these heightened expectations necessitate a proactive approach from the industry, as well as a commitment to invest in technologies and processes that align with the evolving regulatory paradigm. In navigating this complex terrain, stakeholders must not only meet the new requirements but also innovate responsibly, fostering a delicate equilibrium between technological advancement and regulatory compliance. As the draft progresses towards implementation, industry players will need to be agile in adapting to these changes, anticipating the evolving expectations and staying ahead of the curve in the realm of AI-integrated medical devices.

# Conformity Assessment of AI Medical Devices

The draft regulation establishes a comprehensive framework for the conformity assessment of AI medical devices, acknowledging the intricate interplay between emerging technologies and the imperative to uphold safety and efficacy standards. Underpinning this assessment is Article 43, which delineates various conformity assessment procedures specifically tailored for high-risk AI systems.

Providers of high-risk AI systems find themselves at the forefront of this regulatory landscape, tasked with demonstrating adherence to the stringent requirements set forth in the draft. Notably, Article 43 (3) articulates that the conformity assessment for high-risk AI systems falling within the legal acts enumerated in Annex II Section A is governed by those legal acts. This includes both the MDR and IVDR, further integrating AI-specific considerations into the existing regulatory fabric that governs medical devices.

The dual inclusion of the MDR and IVDR within Annex II Section A (under Nos. 11 and 12) ensures a cohesive approach to conformity assessment, mitigating potential overlaps and streamlining procedures. Consequently, providers of AI medical devices must navigate the relevant conformity assessment processes outlined in the MDR and IVDR, ensuring compliance not only with existing medical device regulations but also with the novel requirements introduced by the AI Regulation.

Crucially, the draft empowers Notified Bodies designated under the MDR to extend their purview to the assessment of high-risk AI systems, essentially designating them as authoritative entities—akin to a "Notified Body" under the draft Regulation. This strategic alignment seeks to minimize redundancy in conformity assessments, offering a streamlined approach for AI medical devices that fall under both the MDR and the draft Regulation. However, this convergence raises the importance of ensuring that Notified Bodies possess the requisite internal competence to effectively evaluate tasks performed by external bodies on their behalf.

In certain exceptional cases where Notified Bodies under the MDR or IVDR lack the capacity or expertise required, the potential for a dual referral—one under the existing medical device regulations and another under the draft Regulation—exists. This underscores the necessity for proactive collaboration and communication between regulatory bodies, industry stakeholders, and Notified Bodies to preemptively address any potential capacity gaps and ensure a seamless conformity assessment process.

As the regulatory landscape continues to evolve, all entities involved in the life cycle of AI medical devices—manufacturers, Notified Bodies, regulatory authorities, and end-users—must proactively engage in a coordinated effort to navigate these complexities. The convergence of AI and medical devices demands a harmonized approach to conformity assessment, striking a delicate balance between upholding established safety standards and fostering innovation in this rapidly advancing technological landscape.

# Additional Costs for Industry

The ambitious regulatory framework proposed in the draft AI Regulation is not without its economic implications, particularly for the stakeholders within the medical device industry. As the industry undergoes a transformative shift, embracing the integration of AI into medical devices, a concomitant increase in compliance costs is anticipated. The multifaceted requirements outlined in the draft, spanning risk management systems, testing procedures, technical documentation, transparency, human oversight, accuracy, robustness, and cybersecurity, present a substantial financial burden for manufacturers, distributors, and other economic actors.

According to the Commission's estimations, the cost of compliance for providing an average high-risk AI system, valued at around EUR 170,000, is projected to range between EUR 6,000 and EUR 7,000. These costs, however, represent only the tip of the financial iceberg. In addition to initial compliance expenses, there are ongoing human supervision costs estimated at approximately EUR 5,000 to EUR 8,000 per year. Moreover, suppliers of high-risk AI systems may incur verification costs ranging from EUR 3,000 to EUR 7,500. These figures, while indicative, underscore the substantial financial commitments that industry players must allocate to ensure adherence to the regulatory demands imposed by the draft.

The economic impact is not confined solely to compliance costs; it extends to the broader landscape of research and development. The heightened regulatory expectations necessitate a recalibration of internal processes and product development strategies to align with the evolving compliance requirements. This strategic realignment may introduce additional R&D costs as companies strive to innovate responsibly within the parameters set by the draft Regulation.

While the Commission's estimations provide a preliminary insight, the actual costs remain subject to various factors, including the specific nature of AI applications, the complexity of medical devices, and the evolving technological landscape. As industry players navigate this uncharted territory, a degree of uncertainty persists regarding whether the projected costs will materialize or if higher financial commitments will be required.

The economic considerations are not confined to the immediate stakeholders; they have a trickle-down effect on the end-users as well. The potential for increased compliance costs may influence the pricing of AI-integrated medical devices, posing a challenge for accessibility and affordability. Striking a balance between fostering innovation and ensuring the sustainability of the medical device industry, while safeguarding public health, becomes a delicate task in the face of these economic intricacies.

The draft AI Regulation introduces not only a paradigm shift in the regulatory landscape but also a recalibration of financial commitments for the medical device industry. The true economic impact will unfold as industry players adapt to these regulatory changes, emphasizing the importance of proactive planning, strategic investment, and collaborative efforts to navigate the evolving intersection of AI and medical devices.

# Summary

As the artificial intelligence converges with medical devices, the draft AI Regulation emerges as a landmark initiative by the European Commission. As we navigate the intricate contours of this regulatory framework, it becomes apparent that its far-reaching implications extend beyond legal definitions and compliance procedures. The symbiotic relationship between innovation and regulation, especially in the context of medical devices, presents a delicate balance that stakeholders must strike to ensure the continued progress of AI-driven technologies while safeguarding public health and industry sustainability.

The integration of AI into medical devices promises unparalleled advancements in functionality and diagnostic capabilities, ushering in a new era of healthcare innovation. However, this promise comes hand in hand with heightened expectations for safety, accuracy, and transparency. The draft Regulation sets the stage for a paradigm shift, ushering in a new era where the industry is not only responsible for producing cutting-edge technologies but is also held accountable for the potential risks associated with these innovations.

The comprehensive requirements outlined in the draft, especially for high-risk AI systems, underscore the need for a proactive and collaborative approach. From risk management and testing procedures to transparency and human oversight, the regulatory landscape demands meticulous attention from all involved parties. The alignment with existing regulations, such as the MDR and IVDR, seeks to streamline procedures, but it also requires a meticulous balancing act to ensure a cohesive and efficient conformity assessment process.

Yet, with innovation and regulatory adherence come the economic realities. The projected compliance costs, while estimates, illuminate the financial commitment required from industry players. These costs extend beyond the immediate financial burden, influencing research and development strategies and potentially impacting the pricing and accessibility of AI-integrated medical devices.

The conclusion drawn from the draft AI Regulation is a call to action—a call for industry players, regulatory bodies, and other stakeholders to collaboratively navigate this complex terrain. Striking a harmonious balance between technological advancement, regulatory compliance, and economic sustainability is paramount. As the industry moves forward, adaptability, foresight, and a commitment to responsible innovation will be crucial in ensuring that the promise of AI in medical devices is realized while upholding the highest standards of safety and efficacy. The future of AI in healthcare hinges not only on technological breakthroughs but also on a regulatory ecosystem that fosters trust, transparency, and the collective well-being of society.