The European Union's Artificial Intelligence Act (EU AI Act) will significantly shape the future of AI in healthcare. However, its impact on innovation and competitiveness in the medtech sector remains a topic of heated debate. At the European Health Forum Gastein (September 24-27, Austria), experts discussed how to innovate despite regulation.
To regulate or not to regulate AI
The supporters of the EU AI Act argue that the regulation fosters the sustainable development of trustworthy AI, one that aligns with core European values. Historian Yuval Noah Harari, in his recent book “NEXUS,” compared AI regulation to adding brakes to cars before making them faster - arguing that smart regulation is a prerequisite for responsible innovation.
However, critics point out that Europe risks "regulating first, instead of innovating." They quote the newest European Commission report, “The Future of European Competitiveness” by Mario Draghi, which warns that “the EU is weak in the emerging technologies that will drive future growth” or “Europe is hugely reliant on imports of digital technology.”
“What is the best path forward? Should we first explore AI, invest in its capabilities, and only then regulate, limiting laws to worst-case scenarios? Or should we implement safeguards early in the technology's development?” asked Artur Olesch, moderator of the session organized by the European Coordination Committee of the Radiological, Electromedical and Healthcare IT Industry (COCIR) and Siemens Healthineers.
EU AI Act is now a reality. The focus must shift from asking “if” to understanding “how” the Act will be implemented in healthcare - impacting patients, professionals, and the European medtech industry.
Trust and safety: Core pillars of the AI Act
For Marco Marsella, Director for Digital, EU4Health at the European Commission’s Directorate General for Health and Food Safety (DG SANTE), the AI Act focuses on trust and safety.
"The AI Act goes in the right direction to ensure that the products entering the market are safe for patients," Marsella said, emphasizing the importance of regulation that enhances safety without stifling innovation. He highlighted the role of "sandboxes," which allow member states and stakeholders to test technological solutions before full-scale implementation.
Jelena Malinina, Data Director at Eurordis, added that “from a patient perspective, the Act creates a framework ensuring that products are safe to use,” She cautioned that trust is not a one-time achievement. “Trust is a continuous process,” she said, reinforcing the need for ongoing scrutiny and safeguards.
Dr. Daniel Pinto dos Santos, a radiologist at the University Hospital of Cologne, expressed concerns about over-reliance on AI in clinical practice. "The AI Act suggests that physicians should not place too much trust in AI. Human oversight remains essential," he emphasized. Drawing on his clinical experience, he noted inconsistencies between AI solutions: "We had three different AI vendors for nodule detection in chest X-rays. Despite their CE markings, the results varied widely - even between versions of the same software."
"The AI Act goes in the right direction to ensure that the products that go to the market are safe for the patients," according to Marco Marsella from the European Commission (source: EHFG).
Challenge accepted: Balancing innovation with regulation
Sonja Wehsely, Head of Central Eastern Europe and Central Asia at Siemens Healthineers and Board Member of COCIR, touched on a critical concern shared by many in the healthcare industry: will the AI Act support or hinder innovation? "I hope that in one and a half years, we can say that it has made the industry stronger," she said, "but that depends entirely on how the Act is implemented."
Wehsely highlighted the already significant challenges posed by the Medical Device Regulation (MDR) and In Vitro Diagnostic Medical Devices Regulation (IVDR), which have increased the costs of bringing medical devices to the European market and slowed innovation. "We really need a balanced approach. It’s not just about getting it done for the sake of it; it’s about patients," she stated.
Anni Koubek, founding CEO of QMD Services, commented on these concerns: "The biggest cost from the MDR in maintaining a notified body is always about competence," she said. Koubek stressed the importance of finding a "smart solution" that avoids unnecessary delays. "With AI, we need to ensure we don't add extra time again," she warned.
AI in clinical practice meets real-world challenges
While the AI Act aims to foster innovation, real-world implementation remains challenging. Dr. Santos, reflecting on the current state of AI in healthcare, said that in his experience, the AI solutions on the market today are not necessarily the ones we need most. “They help, but the overall impact is still undecided." He questioned the clinical validation of some AI systems, noting that while they might be technically sound, "we are still missing the key question: does it really help the patient?"
Wehsely shared examples of successful AI applications, including software that significantly reduces MRI scan times. “We recently launched software that can complete an MRI in one-third of the time, improving throughput and potentially reaching more patients. Alternatively, if speed isn't the goal, you can use much less power,” she explained.
Overregulation vs. Transparency – Perceptions of the EU AI Act are diverse, and the legislation stirs strong emotions (source: EHFG)
Low data quality can slow down the implementation of AI
For Malinina, who advocates for rare disease patients, the lack of high-quality data remains a major obstacle. “For rare diseases, the situation is even worse due to data scarcity,” she explained. “Often, patients themselves run the registries, but the data is recorded in inconsistent ways, making it difficult to combine or analyze.”
However, Malinina is optimistic about upcoming initiatives, such as the European Health Data Space. "We have high hopes that a common standard for electronic health records will help improve data quality and encourage more harmonized recording," she said.
The EU AI Act is here to stay. What’s next?
The panelists agreed that the success of the AI Act will depend on its practical implementation, particularly in coordination with existing regulations like the MDR. Koubek noted that notified bodies are ready for the challenge, but legislation must be "realistic and not overly burdensome."
"Some may call it an administrative burden; I would call it transparency," according to Marco Marsella. "When developing AI solutions, especially for healthcare, ensuring that the training data is free from bias is crucial."
Panelists emphasized the need for collaboration in shaping the AI Act’s implementation. Dr. Santos urged stakeholders to keep patient outcomes at the forefront, while Malinina called for "sectoral rules" to tailor AI regulations to specific needs.
Marsella’s final word focused on the balance between safety and innovation. He emphasized that the Act, together with initiatives like the European Health Data Space, will be critical in ensuring that AI improves patient care without compromising safety.
