Algorithmic Opacity, Proxy Discrimination and Financial Citizenship in AI-Driven Credit Markets

Resumo

This article investigates the intersection of artificial intelligence, consumer credit, and financial governance, arguing that algorithmic opacity in AI-driven credit scoring does not eliminate discrimination but transforms it into an epistemic and governance crisis. Methodologically, the research adopts a qualitative, bibliographic-documentary approach grounded in doctrinal legal analysis and comparative law. Data were gathered from legislation, sectoral regulation, peer-reviewed scholarship, and institutional reports produced by the Banco Central do Brasil (BCB), the Bank for International Settlements, the World Bank, and the U.S. Government Accountability Office. The Apple Card controversy is examined as an instrumental case study based on the official report of the New York State Department of Financial Services (2021). The comparative analysis encompasses the regulatory frameworks of Brazil, the European Union, and the United States, allowing for the identification of structural convergences and divergences in institutional responses to algorithmic opacity. The study demonstrates that historical biases are statistically reconstructed through proxy variables while the inherent opacity of machine learning models conceals these processes under a veneer of data-driven objectivity. The result is a structural deficit in evidence, contestability, and supervisory capacity that undermines the core pillars of financial citizenship: protection, participation, and trust. The original contribution lies in reconceptualizing algorithmic credit discrimination as a problem produced by opacity rather than solely by intentional disparate treatment, revealing that the absence of evidence of bias cannot be equated with evidence of fairness when the system under scrutiny is designed in ways that resist independent verification