Erros de sistemas de IA e a responsabilidade das empresas desenvolvedoras no ordenamento jurídico brasileiro

Abstract

This paper aims to analyze the civil liability of companies that develop generative artificial intelligence systems in cases of errors, failures, and damages caused to users and third parties under the Brazilian legal system. The research is based on the understanding that the evolution of the media, the internet, and algorithms has profoundly changed the way information is produced and disseminated, leading to the emergence of systems capable of creating texts, images, audio, and synthetic content with an appearance of truthfulness. In this context, algorithmic opacity, the difficulty of identifying causation, and the multiplicity of agents involved in the creation, provision, and use of these technologies challenge traditional categories of civil liability. The study examines the application of strict liability, the risk-profit theory, the Brazilian Civil Rights Framework for the Internet, the General Data Protection Law, and the recent interpretation of the Federal Supreme Court on the Themes 987 and 533. It also analyzes paradigmatic cases involving systems such as Grok, Microsoft Tay, and Chai App “Eliza”, as well as the risks posed by artificial intelligence to democratic integrity, especially in the electoral context. Finally, the paper discusses containment and regulatory mechanisms, with emphasis on Bill No. 2,338/2023, Bill No. 2,630/2020, and algorithmic accountability. The research concludes that Brazilian law already provides relevant instruments for liability and damage prevention, but still requires specific regulatory improvement to address the particular risks of generative artificial intelligence, ensuring transparency, governance, adequate compensation, and protection of human dignity.