Korišćenje žigom zaštićenih oznaka prilikom „treniranja“ naprednih sistema veštačke inteligencije

Given the increasing use of advanced artificial intelligence systems in products that are part of individuals’ daily lives, the professional community’s attention is particularly focused on the challenges regarding the legal protection of the information utilized by these systems. Not every information enjoys the same level of legal protection – the legal regime of protection depends on whether the information is personal data, a trade secret, an intellectual property protected by intellectual property rights, or otherwise. The authors’ focus is on the use of trademark-protected marks in the process of training advanced artificial intelligence systems – this is the scope of the research presented in the paper. The initial thesis posits that the extent of mark protection is determined by its primary function of distinguishing products in commerce. Therefore, it is examined whether and to what extent this function is disrupted during the training of artificial intelligence systems. In an effort to define the boundaries of trademark protection in this context, the author analyzes the rights and obligations of various parties involved in this process – those who train AI systems and market the products based on them, the rights holders of the marks used for training, and the users of such products. The research is limited to the domestic legal framework, with a reference to normative solutions within the European Union. Finally, the author examines whether certain rules on intermediary liability for trademark infringement can be applied in this field, considering existing practices and legal solutions.