AUSTRIA Law and Practice Contributed by: Harald Strahberger and Florian Sesztak, Kinstellar
There are no specific safe harbour provisions in Aus - trian law regarding AI, but courts generally assess the reasonableness of conduct. For example, if an AI sys - tem provides a false recommendation due to internal algorithmic bias or flawed training data, a provider who mindlessly relies on the system without verifica - tion may still be held responsible. A unique challenge arises from the “black box” prob - lem, where the internal functioning of AI is non-trans - parent. In such cases, Austrian courts may apply a burden-shifting approach, requiring the developer or vendor to prove that the system performed correctly. Transparency and explainability will play a growing role in liability assessments going forward. 5. Ethical and Governance Considerations for Healthcare AI 5.1 Ethical Frameworks Austria does not have a single binding ethical code specific to healthcare AI, but developers and institu - tions operate under a combination of EU-level ethical guidelines, national laws, and institutional codes of conduct. The EU High-Level Expert Group on AI has issued the Ethics Guidelines for Trustworthy AI, which Austria has adopted as a voluntary standard, espe - cially in public research settings. These guidelines emphasise core principles such as human autonomy, prevention of harm, fairness, and explicability. While not legally binding, these principles influence the regulatory implementation, which will codify many ethical principles into enforceable obligations for high- risk AI systems. In Austria, ethical considerations are also addressed at the institutional level (eg, university hospitals and ethics committees), particularly when health data is processed or AI tools are deployed in clinical trials or patient care. 5.2 Transparency and Explainability Transparency and explainability are essential compo - nents of healthcare AI compliance in Austria. Under Articles 12, 13, 14, and 15 of the GDPR, patients have the right to know how their data is used, includ - ing when automated decision-making or profiling is involved. Where AI contributes to patient care, health -
care providers must ensure patients are informed – especially if the AI influences diagnosis or treatment decisions. The MDR further requires that medical devices, includ - ing AI software, be accompanied by instructions for use that enable clinicians to understand and properly apply the technology. With the AI Act, transparency will become mandatory for high-risk AI systems. Developers will be required to provide documentation on system logic, capabilities, limitations, and risks, not only for regulators, but also for end users such as clinicians and, where relevant, patients. In clinical practice, Austrian institutions will have to implement disclosure procedures for AI-assisted diagnostics and decision support, though this is not yet standardised. Nonetheless, regulatory and ethi - cal norms suggest a duty to inform and educate both professionals and patients about the role of AI in care decisions. 5.3 Bias and Fairness Although Austrian laws do not contain obligations to test or mitigate algorithmic bias, this issue is increas - ingly addressed under GDPR, MDR, and the AI Act. • Under Article 5 (1)(a) GDPR, data must be pro - cessed fairly and lawfully, which implies the need to address bias where it may lead to discrimina - tory effects. Moreover, cautions against automated processing that results in unjustified discrimination, especially in health contexts. • The MDR requires clinical evaluation of perfor - mance across the intended user and patient popu - lation, and this implicitly demands attention to data representativeness and bias risks. • The AI Act requires bias mitigation strategies, documentation of training data diversity, and post- market monitoring of AI performance — especially for high-risk systems used in healthcare. In addition, Austrian ethics boards and research funders are increasingly expecting applicants to address equity, demographic fairness, and the pro - tection of vulnerable groups.
24
CHAMBERS.COM
Powered by FlippingBook