A new framework for European Responsible AI Development (EURAID), designed to help German hospitals safely develop and implement artificial intelligence systems, has been published in the Journal of Medical Internet Research.
Chief Medical Officer Eva Weicken and Dr. Katharina Weitz from the Applied Machine Learning group in Fraunhofer HHI’s AI Department were part of the interdisciplinary team that developed the framework, contributing expertise in AI evaluation and responsible, human-centred AI design for healthcare.
“The development and implementation of AI systems into the complex clinical routine of a hospital is often challenging in practice. With EURAID, we have developed practical, human-centred guidelines that engage all relevant stakeholders from the start and aim to enable responsible and effective AI use in hospitals,” says Eva Weicken.
Practical guidance for “in-house” AI development
Artificial intelligence is advancing rapidly, yet its integration into everyday clinical practice remains cautious and uneven. Many promising systems fail to integrate into routine care because they are developed and tested without sufficient consideration of hospital-specific workflows, governance structures, and the needs of healthcare professionals. As a result, systems are often perceived as an additional burden rather than support.
The EURAID framework aims to address this challenge by outlining how hospitals can develop, validate, and deploy AI systems “in-house”, ensuring solutions are tailored to local clinical needs and datasets while remaining aligned with the EU’s evolving regulatory landscape.
Recommendations include:
● establishing cross-functional governance structures
● clarifying ethical, legal, clinical, and technical responsibilities
● assessing institutional readiness for AI deployment
● integrating iterative testing and evaluation processes
● ensuring transparency, safety, and human oversight
The recommendations draw on insights from multi-stakeholder workshops involving hospitals, regulators, insurers, labour unions, and research institutions.
Human-centric AI in healthcare
A central principle of EURAID is that healthcare professionals should be involved throughout the entire lifecycle of an AI system. Rather than treating clinicians solely as end users, the framework positions them as co-designers, implementers, and evaluators of AI systems. This helps ensure that new tools genuinely support clinical practice and patient care rather than creating additional workload or operational risks.
“We need both: involving clinicians in the development process and supporting them in building AI literacy so that these new tools fit clinical practice, support their work, and ultimately benefit patients,” says Dr. Katharina Weitz.
An interdisciplinary collaboration
Led by Prof. Stephen Gilbert from the Else Kröner Fresenius Center for Digital Health at TU Dresden and Dr. Anke Diehl from University Hospital Essen, the framework was a multidisciplinary collaboration involving experts in medicine, hospital management, medical informatics, quality management, occupational safety, law, ethics, and human-centred AI. Anett Schönfelder is the first author of the publication and a researcher in Prof. Stephen Gilbert's team.
Partners include the Karlsruhe Institute of Technology (KIT), the German Federal Ministry of Health (BMG), the Federal Institute for Occupational Safety and Health (BAuA), health and accident insurers (AOK and DGUV), labour unions (ver.di), and additional stakeholders from policy and academia.
The publication has already received attention in German health policy and clinical media, including coverage in Der Tagesspiegel and Deutsches Ärzteblatt.
A German-language summary of the framework has also been published in the BAuA “Bericht kompakt” series by the Federal Institute for Occupational Safety and Health (BAuA). Read it here.