A ilusão da anonimização e os riscos de discriminação: dados sensíveis e o uso de IA sob a perspectiva da LGPD

Abstract

This research is based on a rigorous legal analysis of the use of sensitive data by artificial intelligence (AI) systems. The topic, which is highly relevant in the context of contemporary digital law, addresses technological expansion and the need to assess the adaptability of the General Data Protection Law (Law No. 13,709/2018 – LGPD). The central problem driving this investigation can be formulated as follows: to what extent is the anonymization provided for in the LGPD effective in protecting sensitive data against discrimination in artificial intelligence systems, and what legal and technological adaptations are necessary to mitigate these risks? That said, the main hypothesis underlying this work is that anonymization, as outlined in the LGPD, is not sufficient to eliminate the risks of discrimination that accompany the use of sensitive data by artificial intelligence algorithms. This is due, with all due respect, to the latent possibility of re-identification of previously anonymized data and the manifest presence of algorithmic biases in AI systems, which can generate discrimination that is not fully mitigated by current legal provisions. Furthermore, there is a hypothesis that the absence of effective oversight and specific regulations for the use of AI in conjunction with sensitive data contributes unequivocally to the perpetuation of these vulnerabilities in the Brazilian legal system, which lies in the urgency and exponential nature of the use of AI in vital sectors, such as health, education, and the labor market, where automated decisions directly impact fundamental rights, especially the right to equality and non-discrimination. The study will thus seek to contribute to a critical understanding of the gaps in legislation and suggest solutions that strengthen protection against algorithmic discrimination. Ultimately, it is considered that proposing technical and legal solutions could mitigate the risks of discrimination, promoting greater compatibility between the LGPD and contemporary technological practices