Patientskadeersättning i den artificiella intelligensens tidsålder - En rättsdogmatisk undersökning av säkerhetskrav, rättsliga utmaningar och reformbehov vid användningen av AI-baserade medicintekniska produkter

Sammanfattning

The increasing use of artificial intelligence in healthcare raises several legal issues. One of those is how patient injury compensation should be handled in case of patient injuries caused by AI-based medical devices (MTP), for example, when they are used to analyze mammography images. Although AI technology has demonstrated a satisfactory level of ability to identify signs of breast cancer, errors will inevitably occur at some point. When this happens and a patient is misdiagnosed, the legal system needs to be clear on how to deal with the damage. The purpose of this thesis is to provide clarity on how patient injuries caused by AI-based MTP should be managed legally. It focuses on situations where AI is used to analyze mammography images for signs of breast cancer. The thesis is based on a legal dogmatic method and what I have chosen to call the EU legal method. The study is conducted from a patient perspective, with the injured patients' interests at the center. The first question of the essay explores which additional requirements for AIbased MTP (specifically SaMD) the AI regulation adds in addition to the requirements already imposed by the Medical Device Regulation (MDR). The analysis shows that the adoption of the AI-regulation adds requirements, for among other things, transparency, logging, and human oversight. The new requirements are linked to the specific risks posed by the AI technology. The second question concerns challenges that arise when the current patient compensation rules are applied to cases of injuries caused by AI-based medical devices. It can be concluded that there are several legal challenges for the injured patient. The difficulties are mainly linked to the burden of proof. The low transparency of AI systems, the “black-box problem”, renders it difficult for the injured party to prove the existence of a defect in the AIbased medical device and that the defect caused the injury. The third question explores what changes to the Swedish Patient Injury Act [patientskadelagen] could be considered to better address injuries caused by AI-based MPT. Inspired by the now withdrawn proposal for the AI-liability directive (AILD) and the revised Product Liability Directive (PLD), a special rule of presumption is proposed. In addition, it is proposed that the court should be able to order the manufacturer of an AI-based MTP to provide logs and technical documentation kept following the AI regulation, that can be used by the injured party to prove fault or causation. I also suggest that the “experienced specialist” [erfaren specialist] standard used in the assessment of diagnostic injuries should be revised to reflect the fact that it is no longer only human specialists who analyze medical data like for example mammography images.