Modelos generativos e mecanismos de tutela da pessoa humana: regimes de responsabilização de agentes autônomos perante a regulamentação da IA de propósito geral no ordenamento jurídico brasileiro

Abstract

The essay addresses the need for clear and effective regulation of generative artificial intelligence in Brazil, especially in relation to Bill No. 2,338/2023. The study aims to analyze the legal implications of generative AI, highlighting the importance of legislative adjustments that respect human rights and ensure safe and reliable AI. The research also explores the feasibility of applying the concepts of legal personality and civil liability, according to the theories of Beckers and Teubner, to generative models. Furthermore, while respecting the particularities that differentiate German private law from Brazilian law, it seeks to investigate the applicability of this theory within the Brazilian legal context. Despite adopting the authors as theoretical references, this dissertation does not intend to use comparative law. The methodology employed is the deductive method with a qualitative analysis, aiming to understand the dynamics of social relations and the legal precepts that are being defined. The research is exploratory, given the still-developing nature of AI legislation in Brazil, and focuses on the specific implications of generative AI, which brings unique and complex issues. Finally, the work explores the need for a legislative framework that allows the Brazilian legal system to adapt to the innovations and conflicts imposed by generative AI, avoiding ambiguities that could compromise legal certainty and ethics in the development and use of this technology. The present study was developed within the scope of the author's participation in the Núcleo de Estudos Avançados em Pessoa, Inovação e Direito (NEAPID) at the Law School of the Federal University of Juiz de Fora, as well as in the research project "Inovação e Direito na IA: mapeamento normative e análise do exercício de direitos fundamentais" (CNPq-universal), coordinated by Professor Sergio Negri, the supervisor of this work.