PRESS RELEASE: EUHA Issues a Position Paper on Artificial Intelligence

Federated Data Handling: EUHA supports federated, hospital-level data environments rather than centralised national warehouses. This model ensures GDPR compliance, avoids complexity, and enables cross-border research. To address inconsistent GDPR interpretations, EUHA proposes an EU-wide network of accredited healthcare Data Protection Boards, alongside EU-level directives and funding to secure interoperable and FAIR data standards (Findable, Accessible, Interoperable, and Reusable).

Data Custodianship by Healthcare Providers: Healthcare providers must remain the primary custodians of patient data. EUHA’s position is that research in the public interest constitutes a legal basis for the secondary use of data by public healthcare providers or universities. We acknowledge that research done in private institutions can also be in the public interest.  This should be combined with a broad opt-out model. Patients deserve clear, standardised information on how their data and AI tools are used. Providers, who bear the burden of data preparation and validation, should be recognised and compensated for this responsibility.

Health and Wellness Data: The separation of health and lifestyle data is artificial. Behavioural and wellness data, such as those from diet or exercise apps, can directly impact health and must be regulated to the same standards as clinical data. EUHA calls for a new category of health-related data that meets GDPR, Medical Device Regulation (MDR), quality, and interoperability requirements to support preventive and integrated care.

Continuous Validation of AI: AI tools must be validated not only before release but throughout their use in clinical practice. EUHA proposes conditional approvals tied to robust post-market surveillance using real-world data in secure hospital-based environments. Continuous monitoring will help detect bias, ensure safety, and maintain trust. Validation outcomes should be publicly accessible, enabling informed decisions by clinicians and patients.

Implementation in Clinical Practice: AI in healthcare must follow established benefit–risk principles, with patients involved in shared decision-making. For black-box systems, individual caregivers cannot bear full liability; EUHA supports a no-fault liability model at the EU level. To prepare the workforce, AI training should be embedded across medical education and professional development, coordinated at the European level to build a digitally prepared healthcare sector.