AI regulation: Association for Digital Healthcare calls for a special approach
To bring AI applications in medicine to market faster and more safely, the industry association SVDGV has its own "sandbox".
Health, data, health data, register data
(Image: Tex vector/Shutterstock.com)
The German Association for Digital Healthcare (SVDGV) has presented a “Green Paper” to break the bottleneck of unclear regulations, a lack of data access, and a lack of test environments for AI developers. In it, the association, which represents over 170 e-health companies, calls for an industry-specific “AI Regulatory Sandbox” for the healthcare sector. The aim is to accelerate market access for medical AI applications, safeguard patient safety, and support small and medium-sized enterprises (SMEs) in particular.
Double regulation and a lack of data as obstacles
According to the association, AI innovators in the healthcare sector face massive hurdles. The core problem is the regulatory uncertainty caused by the threat of double regulation: AI-based medical devices are subject to both the strict Medical Device Regulation (MDR) and the new EU AI Act. How these two sets of regulations interact is unclear in many areas.
The situation is particularly critical for AI systems that continue to learn after their market launch, as there is currently no “clear legal framework” for their authorization and monitoring. In addition, access to high-quality data for training and validation is difficult. While the Health Research Data Centre (FDZ) is still being established, there is a lack of practical, annotated data sets and, above all, machine-readable medical guidelines. The SVDGV sees a further bottleneck in the fact that “test environments […] have so far focused heavily on clinical (inpatient) settings, while outpatient validation has been ignored.”
Videos by heise
The sandbox as a “medical AI agency”
At the heart of the SVDGV proposal is an “AI Regulatory Sandbox,” which is intended to be much more than just a testing ground. The proposal is for a type of “Medical AI Agency” that fulfills two central tasks:
Service hub for companies: It would act as a service provider and provide targeted support to innovators, especially SMEs. This includes advice on classification and study protocols, access to test data sets, and support from an ethics committee specializing in AI.
At the same time, the sandbox is intended to be a state-moderated space in which regulatory authorities, designated bodies, and scientists can jointly clarify open legal issues. Standards for dealing with general-purpose AI (GPAI), the definition of "human oversight," or the harmonization of the different risk assessments of the MDR and AI Act are to be tested in practice here.
To prevent high costs and bureaucracy from stifling innovation, small and medium-sized enterprises are to be given preferential, cost-reduced, and unbureaucratic access to the sandbox. The sandbox should also make standardized templates for documentation, guidelines for regulatory processes, and technical benchmarks freely available as “public goods.”
At the end of a successful sandbox process, there should be final reports. These “exit reports” should bundle the regulatory, clinical, and economic results and serve as a recognized basis for the subsequent conformity assessment and reimbursement by the health insurance companies.
Open questions
“With the AI Regulatory Sandbox, we are building a bridge between regulation, science, and healthcare,” explains SVDGV board member Anisa Idris. In this way, Germany could become a pioneer for the safe use of AI in medicine.
However, the road to this goal is paved with numerous unanswered questions, which the Green Paper itself raises: How will the sandbox be funded? Who will bear responsibility, control, and supervision, and how will it be ensured that the procedures remain truly lean? There is also the question of how legally binding the results of the sandbox will be.
With its proposals, the SVDGV is directly opposing the current plans of the Federal Network Agency (BNetzA), which is legally obliged to set up at least one cross-industry AI real-world laboratory and has already started a pilot project for this purpose. With the Green Paper, however, the healthcare industry is formulating a clear counter-position: instead of a standardized solution for all sectors, it is calling for a specific infrastructure for the highly regulated healthcare sector that meets the special requirements for patient safety and clinical evidence.
(mack)
