Entre riscos e algoritmos: uma proposta ética e prospectiva para os possíveis danos provocados por sistemas de inteligência artificial e perspectivas regulatórias

Abstract

Artificial intelligence systems equipped with self-learning mechanisms and deep learning neural networks, especially after the popularization of generative systems, have been increasingly dominating diverse spheres of work, health, leisure, and human development. Based on specific input data, algorithms dictate the rules of learning and interaction with the environment and, from that point onward, it may no longer be possible to control output data or the potential risks the system may present. Moreover, artificial intelligence systems are endowed with mechanisms of continuous self-improvement, and it is precisely at this point that the central question of this work arises: would it be possible to apply the notion of development risks to defects displayed by artificial intelligence systems? Could suppliers of artificial intelligence systems be held liable for risks not yet known, arising from the self-learning (machine learning) and deep learning characteristics of these systems? The research hypothesis is grounded on the principle of safety as an essential value for any society and as a cornerstone of the liability framework established by the Brazilian Consumer Defense Code, thereby ensuring a more solid basis of trust and reducing complexities in the field of tort law. In a context marked by excessive enthusiasm surrounding automation and the acceleration of processes, as well as the algorithmic optimization of profitable outcomes, it is imperative to reflect on liability for damages. Accordingly, this study begins with an overview of artificial intelligence and its relevance in the era of the Fourth Industrial Revolution, in order to map certain abusive practices and possible unjust harms that justify the need to protect consumers in an era of artificial intelligence characterized by risks and uncertainties. It then proceeds to analyze the theory of development risk and its relationship with autonomous systems endowed with artificial intelligence. The general objective of the research is to focus on potential violations of fundamental rights caused by artificial intelligence systems equipped with self-learning models, and to analyze the feasibility of applying legal instruments already available in the Brazilian legal system to prevent unjust harm. More specifically, it explores the sustainability of a proactive liability model and risk management approach. The research employs the hypothetical-deductive method and theoretical-bibliographic investigation. In the end, the concluding remarks will present an overall panorama of the investigation, at which point the study aims to clarify the results achieved through the path taken by the research.