Összefoglaló. A tanulmány kezdő axiómája a mesterséges intelligencia biztonságos alkalmazása. A biztonságos alkalmazás egyik aspektusa a jogi biztonság, az a jogi környezet, amelyben a felmerülő jogi kérdések rendezésére alkalmazható keretrendszer áll rendelkezésre. A tanulmány a Semmelweis Egyetem projektjében fejlesztett mesterséges intelligencia alkalmazásának olyan polgári jogi problémáit vizsgálja, amelyek a mindennapi hasznosítás során merülhetnek fel. A tanulmány következtetése szerint a vizsgált mesterséges intelligencia szerzői műnek minősül és több védelmi forma is alkalmazható. A jogi szabályozás de lege ferenda kiegészítésre szorul a szerzői mű folyamatos változása okán. Szükséges rögzíteni egy referenciapontot, amely a felelősség kiindulópontjául szolgál.

Summary. The starting point of the study is the safe use of artificial intelligence. Legal certainty is one aspect of safe usage, the legal environment in which a framework is available that can be used to resolve legal issues. The paper examines the civil law issues that may arise in the everyday use of the artificial intelligence application developed within the Semmelweis University project. The study will first focus on the legal protection of the Semmelweis AI, including whether this protection is currently international, regional (European Union) or national and which of these is the optimal choice. The study also reflects on the legislative preparatory work of the European Union in this regard. Our hypothesis is that the majority of civil law areas concerning AI can be regulated within a contractual framework. The AI software developed by the project is a forward-looking medical and practical solution. If we want to use a legal analogy, we can imagine its operation as if we had a solution that could analyse all the national court decisions in each legal field and provide an answer to the legal problem at hand, while simultaneously learning and applying the latest court decisions every day. For this AI solution, the diagnostic process must be carefully examined in order to identify the legal problems. I believe that the optimal solution is to classify this AI application as ‘software’ because this allows property rights to be acquired in their entirety and it opens the door to clarifying individual associated usage and copyright by contract. An important civil law question arises in relation to parallel copyright protection, when the individual personal contributions (creative development work) to the software cannot be separated. Therefore, it is important to record the process and to separate the individual contributions protecting by copyright. The AI plays a questionable role in the diagnostic process. If the software itself cannot make a decision, but only provides a framework and platform, then it will not be entitled to co-ownership relating to the diagnostic images (e.g. just as a camera will not own the rights to the pictures taken with it). However, if the algorithm is part of the decision-making (e.g. the selecting of negative diagnoses), it would possibly be co-owner of the right, because it was involved in the development of the classification. All this should be clearly stated in the licence agreement, based on full knowledge of the decision-making process. However, de lege ferenda, the legal regime needs to be supplemented in view of the constant changes of the copyright work and the changing authors. There is a need to establish a specific point in the legislation that serves as a reference point for liability and legal protection. The issues under consideration are of a legal security nature, since without precise legal protection both the creator of artificial intelligence and the persons who may be held liable in the event of a malfunctioning of such systems may be uncertain.