OS PARADOXOS DA DEMOCRATIZAÇÃO DAS IAS GENERATIVAS DE IMAGEM: EM BUSCA DE UMA TUTELA JURÍDICA FUNDAMENTADA NO GRAU DE CONTROLE DO PROCESSO CRIATIVO A PARTIR DO FUNCIONAMENTO DOS MODELOS DE DIFUSAO LATENTE

Abstract

Artificial Intelligence (AI) has generated tensions in the most diverse fields of legal knowledge. In the field of Copyright, it is seen as a threat by some sectors. The lack of parameters for assessing the originality of AI creations has raised debates about the need to improve the structures that underpin the recognition and effective protection of intellectual works. Given this scenario, the research problem of this work is: what parameters should Copyright Law adopt to deal with the challenges posed by generative artificial intelligences in Brazil, guaranteeing the protection of authors' rights while promoting technological progress and democratic access to content creation? The general objective is to provide subsidies for the development of a new Brazilian Copyright Law format that is equipped to deal with these challenges from a perspective that embraces the new art forms, ensuring the protection of authors' rights while promoting the democratization of access to content creation and encouraging ethical technological advancement. As a result, the aim is to demonstrate that the outputs of diffusion models are original creations and should therefore be covered by copyright. The analysis of the legal challenges presented by latent diffusion models - including datasets, the models themselves, refinement and outputs - highlights the need for legal protection based on the degree of control exercised by the user over the creative process. The final conclusion is that the concept of originality is often ignored in legal systems such as Ukraine, the United States and Brazil itself, on the assumption that these models are not capable of generating novelty. This scenario repeats a historical stumbling block of Copyright Law, observed since its creation and with the advent of the internet, delaying technological development and democratization. It also invalidates a new way of making art, syntography. The paper highlights the need to update laws to take account of these innovations, especially in the face of the inevitability of design software equipped with powerful diffusion models and a new generation that interacts daily with neural networks. The methodology adopted is of an explanatory nature, with a comprehensive review of existing literature and analysis of concrete cases of conflicts involving copyright and the use of artificial intelligence in the creation of artistic works. Solutions adopted in other legal systems, such as Europe, Ukraine and the United States, were examined in order to identify practices that could inspire Brazilian legislation. This study contributes to the debate on the relationship between generative artificial intelligences and Copyright Law, proposing possible ways to deal with the challenges presented by the advance of these technologies in the legal field. It emphasizes the importance of legislation that balances the protection of authors' rights with encouraging technological progress and the democratization of content creation.