Brazil confronts AI‑driven Pix fraud surge and judicial automation

Reports from São Paulo show that cases of Pix fraud have more than doubled since generative AI chatbots became widely available at the end of 2022. The number of estelionato (fraud) decisions by the state court rose from 1,073 in 2024 to 2,270 in 2025, with prosecutors linking the rise to AI‑generated content such as deep‑fake videos, voice cloning and mass‑produced phishing messages. Fraudsters also use AI to evade facial‑recognition tools and craft convincing romance‑scam narratives. The investigation success rate remains low, with the Brazilian Federal Police solving only about 2 % of such cases.

At the same time, researchers at the University of São Paulo’s Labic are piloting AI tools to automate legal tasks. The Agents4Gov program, coordinated by Ricardo Marcacini, aims to provide secure, privacy‑preserving AI for public‑sector governance. Its LLM4Gov framework, already deployed by the Procuradoria Geral da Fazenda Nacional, anonymizes sensitive data, uses a teacher‑student model to train specialized language agents, and complies with Brazil’s LGPD and the EU’s GDPR. Officials say the system can generate detailed reports for managers and help citizens hold public officials accountable.