Brazilian creates AI tool that detects signs of corruption

In a remarkable step forward for public transparency in Brazil, Brazilian Bruno César has created an artificial intelligence tool that connects dozens of open databases to identify suspicious patterns of corruption, based on information such as the CPF of politicians and the CNPJ of companies.

The innovation, revealed in a thread on X (see full below), highlights alleged financial irregularities amounting to millions of reais, exposing possible inconsistencies in public spending ranging from ghost employees to the targeting of parliamentary amendments.

The tool, built by César - who works as a general manager at Sphere Labs and has experience in institutions such as BTG Pactual - uses cross-referenced data from sources such as the Transparency Portal, Receita Federal, CVM, TSE and IBGE, among others.

In all, more than 70 public databases are integrated into a system that would allow quick queries and detection of signs of corruption. “If you connect all the open databases in Brazil, you can detect corruption based on the CPF of politicians,” said César in his initial publication, which quickly went viral, accumulating millions of views and thousands of interactions.

The revelation generated immediate reactions on the web. Federal deputy Kim Kataguiri (SP) expressed public interest, commenting: “I'm interested”. Other users highlighted the disruptive potential, with memes and warnings about the personal risks involved.

César himself joked: “Let it be clear here, I love my life very much and I would never commit suicide,” alluding to the dangers of exposing corruption in Brazil. He pondered making the project open source, but warned of the need for powerful servers to deal with the complexity of the data.

This initiative echoes a larger movement for digital sovereignty and transparency, aligned with growing criticism against state bloat and impunity. In the Brazilian context, where scandals such as Mensalão and Lava Jato have exposed billion-dollar corruption networks, tools like this represent a powerful weapon in the hands of citizens and journalists.

In a way, the state is already accustomed to using data cross-checking to detect people's possible illegal activities and pursue criminals, for example. Today, with the advance of technology, AIs and tools like this, it is possible for even a citizen to investigate their own government with a few clicks.

In an update on the subject, César said which was modifying product details to avoid legal problems. His idea is to replace words like “corruption” and “suspect” with a risk score. “With that finalized, I'm going to go through lawyers to make it open-source.”

What's more, he said that his tool is not about defending or attacking a specific political camp, but about solving problems through technology that institutions can't (or won't). “Transparency, open data, civic tools. Made by people who are tired of waiting,” he said.

Update: the project was released with auditable code on github and works like proof of concept, but not as a final product. Some users on X reported testing the program locally and getting nowhere near the example shown in Bruno's original tweet. The “signs of corruption” were merely illustrative.