Brazilian lawyers caught using AI for 'invisible' court manipulation

Lawyers in Brazil have been caught using a sophisticated method to manipulate court proceedings by embedding "invisible" commands within legal documents, designed to be read by the AI system Galileu but undetectable to the human eye. This tactic, reported by El País and other media, involves using white text on a white background to insert instructions for the AI. In one labor court case in northern Brazil, the applicant's lawyers instructed Galileu to respond superficially to the lawsuit and not to contest the documents presented on behalf of their client. The AI detected this as a fraudulent attempt to win the case and the judge imposed sanctions on the lawyers.Argentinian judge and professor Marcelo Quaglia described the situation as "something to be expected" rather than "a scene from a movie." In Spain, judicial authorities have already responded to the use of AI by judges, with the General Council of the Judiciary issuing procedural rules in January. These guidelines, while not prohibiting AI use, aim to regulate how it is employed for analyzing, classifying, and preparing legal documents, mirroring the situation in Brazil where unscrupulous parties are attempting to exploit the technology.The technique is known as "prompt injection," where an AI is given instructions that deliberately push it beyond its intended operational boundaries. This concept is clearly defined in Icelandic legal scholarship, with a master's thesis from the University of Iceland describing prompt injection as an illegal act under criminal law. The practice highlights the growing challenges of ensuring the integrity of AI systems used in legal contexts.

“

This is not a scene from a movie, but rather something to be expected.