Healthcare AI 2025

AUSTRIA Law and Practice Contributed by: Harald Strahberger and Florian Sesztak, Kinstellar

8. Specific Applications of Healthcare AI 8.1 Clinical Decision Support AI-based clinical decision support systems are reg - ulated under MDR as SaMD, typically classified as class IIa or higher (see in detail 2.1 Regulatory Defini - tion and Classification of Healthcare AI ). Under the AI Act, they will also be considered high-risk systems, requiring risk management, human oversight, and documentation of transparency and data governance. Validation requires clinical performance evidence, based on real-world or retrospective datasets. Imple - mentation must ensure traceability, usability, and clini - cian training. Under Austrian law, the restriction that medical services may only be provided by physicians prohibits autonomous diagnostic or therapeutic deci - sions by non-physicians. Clinical decision support systems must therefore function as assistive tools, with physicians retaining full responsibility for their decisions. 8.2 Diagnostic Applications Under the MDR, diagnostic AI tools are classified, based on their intended use and risk, as Class IIa or Class IIb SaMD. They require CE marking, clinical validation, and adherence to ISO standards. The AI Act imposes additional obligations on transparency, traceability, and post-market monitoring. Besides the MDR, MPG and AI Act, there are no specific frame - works, eg, for radiology, pathology, or other diagnostic specialities using AI. 8.3 Therapeutic and Treatment Planning Depending on their clinical impact, AI tools for treat - ment planning are classified as Class IIb or III SaMD. They must demonstrate safety, clinical benefit, and risk control. The AI Act will apply due to their high- risk nature, requiring explainability, human oversight, and documented training data. In Austria, the ÄrzteG strictly limits therapeutic decisions to physicians. Therefore, AI systems may support decision-making but cannot replace or automate clinical judgment. 8.4 Remote Monitoring and Telemedicine AI systems used in remote monitoring and telemedi - cine are regulated under MDR, GDPR, and national

on algorithmic functioning, datasets, and risk assess - ments. Austrian regulatory practice encourages partial disclosure via technical documentation that complies with legal standards without fully exposing proprietary content. 7.3 Ownership of AI Outputs Under Austrian law, outputs generated by AI – such as diagnostic suggestions, image interpretations, or therapy plans – are generally not considered works protected by copyright unless they are sufficiently original. Therefore, such outputs typically do not cre - ate automatic IP rights. Ownership is primarily determined contractually. In academic collaborations, IP generated through joint research is often owned by the institution, but spin- offs or licensing agreements may assign usage rights to commercial partners. 7.4 Licensing and Commercialisation In Austria, the licensing and commercialisation of healthcare AI technologies are governed by general contract principles under the ABGB and the Commer - cial Code ( Unternehmensgesetzbuch – “UGB”). While there is no AI-specific licensing law, agreements must be drafted with sufficient clarity and precision to be enforceable. Common licensing models include SaaS arrange - ments. These licences typically define the scope of use, such as clinical, research, or non-commercial purposes, along with terms covering user access levels, system updates, maintenance, and technical support. Compliance with applicable regulations is increasingly built into contractual frameworks. When AI systems are classified as high-risk under the MDR or AI Act, licensing agreements often include specific obliga - tions regarding post-market monitoring, software version control, and audit readiness. Liability clauses, regulatory warranties, and data protection provisions are also routinely negotiated.

CHAMBERS.COM