Transparência na inteligência artificial: mecanismos de enfrentamento nos projetos de regulação da Europa e do Brasil

Abstract

The aim of this work is to investigate the regulatory documents on artificial intelligence (AI) in Brazil and the European Union (EU) to determine whether the provisions related to transparency and explainability principles consider the technical and systemic characteristics inherent in machine learning-based AI systems, specifically deep neural networks. To achieve this, three theoretical frameworks are established. The first framework addresses the definition of AI concepts, with a focus on machine learning and deep learning, which are essential for a proper assessment of the impact of regulations. The second framework examines transparency as a tool to protect the right to equality, considering private autonomy and the right to information. Finally, the third theoretical framework evaluates the regulatory documents in Brazil and the EU, with a specific focus on transparency and explainability principles