Stanford University School of Law, in the United States, held its LLM Law CodeX 2026 Hackathon, on April 12th. The event sought innovators, machine learning experts and legal professionals to think about the future of technology for Law. Among the participants, there was a company with a Brazilian accent. NL.AI was the only startup from Brazil present. Your solution? A project to detect and map money laundering and corruption schemes.
The tool promises to use artificial intelligence to integrate and analyze data from multiple sources — from public bodies to financial entities — in an automated way. In the prototype presented at the Hackathon, 19 public bases were used.
The solution is based on an AI multi-agent framework. Each of them specializes in its own function: how to read regulatory standards, collect data, reason about patterns, cross-reference sources.
Continues after advertising
Unlike other fraud detection solutions available today, NL.AI’s technology promises to not only generate an alert, but explain why it was issued, which agents participated in the analysis, and what evidence supports the conclusion, producing an end-to-end auditable trail.
“We are talking to 3 of the largest banks in the country. The idea is to test our technology within them. These are proofs of concept”, says the chief researcher of AI strategies at NL.AI, Ricardo Fernandes. Companies in the financial system are the startup’s main focus.
Fernandes is a postdoctoral fellow in legal artificial intelligence at CodeX with a focus on AI and innovative technologies applied to Law. The development of the startup is led by him together with Helano Matos, he has a doctorate from the University of Liverpool and a postdoctoral degree from King’s College London.
However, the company has not yet started operating. The expectation is that the business will be running within 12 months.
Among the bases analyzed by NL.AI are regulatory references such as Central Bank and Coaf standards and typologies; Federal Revenue registration and corporate bases; lists of politically exposed people and electoral donation records; Official journals and cross-references of government acts; transparency portals; legal proceedings for civil, administrative or criminal offenses; sanctioning information, such as international sanctions lists and national registers of punished companies; and, also, government transparency data, including public contracts and resource transfers.
The Stanford LLM Law Hackathon is part of broader programming at CodeX, a legal innovation center at Stanford University that brings together Law, Engineering and Business professionals and students to develop AI-based solutions applied to the legal sector. Despite participation, NL.AI did not win.
