The European Union is advancing artificial intelligence regulations that could impact medical device and diagnostic companies around the world. Through the policy, the EU plans to impose additional requirements on the use of AI in medtech and issue fines for noncompliance that could total billions of dollars.
In a document setting out its plans, the European Commission states high-risk AI systems should only be placed on the EU market "if they comply with certain mandatory requirements" that ensure they do not pose unacceptable risks to important public interests. Medical devices and in vitro diagnostics are on a long list of products that could be classed as high risk.
The proposal acknowledges the risk of overlaps with other regulations and the potential to impose unnecessary burdens. Citing medical devices as an area in which overlaps may occur, the EU said the existing conformity assessment procedures will be used to check the proposed AI requirements.
Yiannos Tolias, a lawyer at the Commission’s health group, set out the need for AI requirements in addition to the existing Medical Device Regulation at an event held before the publication of the plans.
"The new horizontal AI proposal would come to complement or to stand next to the MDR," Tolias said. "An AI medical device ... would be now more secure, in the sense that it will be also complying with the MDR obligations and in addition those aspects of AI that could be creating some worries and some concerns would be handled by the new AI regulations. So, the two would be ensuring that the system is secure and trustworthy and so on."
Large penalties await companies that fail to comply with the EU requirements on AI. The EU plans to issue fines of up to €30 million ($36 million) or, if the offender is a company, 6% of its total worldwide annual turnover for infringing the rules.
If applied to all of Johnson & Johnson, not just its device group, a 6% fine would equal almost $5 billion, based on its 2020 turnover. For Medtronic, 6% of 2020 turnover equals $1.7 billion. In both cases, the fine would put a significant dent in earnings.
Researchers have raised concerns about the EU proposal. In a blog post titled "EU proposing to regulate the use of Bayesian estimation," Bob Carpenter, senior research scientist at the Flatiron Institute's Center for Computational Mathematics, said the AI definition "feels hopelessly vague with phrasing like 'wide variety of methods' and the inclusion of 'statistical approaches'."
There is precedent for the EU being perceived, in some quarters at least, to create regulations that are vague, complex and ultimately do more harm than good. The phrase "bureaucratic overreach" has been used in relation to MDR, the clinical trials directive was slammed for hindering access to new therapies and Eli Lilly called REACH a "substantial regulatory and financial burden" that has a "significant detrimental impact on the competitiveness of pharmaceutical manufacturing."
The impact of other EU regulations extends beyond Europe, with some laws becoming the de facto standard overseas. As the General Data Protection Regulation deadline approached in 2018, Forbes reported U.S. businesses had spent $7.8 billion on preparing for the changes. There are signs the Commission sees AI regulations having similarly far-reaching influence, for example in its statement that the proposal "positions Europe to play a leading role globally."
If the regulation stymies the advance of AI in healthcare, Europe could suffer more than many other regions. Last month, EIT Health, a group funded by the EU, published a report that set out the health challenges facing the region, explaining that one in four Europeans are predicted to be aged 65 years or older by 2050. AI could help manage the change by making healthcare more efficient.
EIT Health contends that changes are needed if the EU is to realize those benefits. The group said the current approach is too geographically fragmented, leading it to propose the creation of "a more robust data infrastructure for Europe" to "ensure member states and healthcare systems can share data, allowing experts to track diseases, diagnose more quickly and develop new and improved AI-based solutions." EIT Health published its report days before the Commission shared its EU-wide proposals.
The conclusions reached by EIT Health are in line with some of the findings of a 2020 report by trade group MedTech Europe and Deloitte, which cited the "fragmented data landscape" and the "different legal frameworks regulating AI and data in healthcare" as barriers to the realization of the potential of AI. MedTech Europe calculated clearing the barriers could save 400,000 lives a year and free up labor equal to the work time of 500,000 healthcare professionals.