Brazilian lawyers fined for attempting to manipulate court AI
Two lawyers in Brazil have been fined $16,000 after embedding hidden prompts in legal filings to influence the court’s artificial intelligence system.
A Brazilian labor court has imposed fines totaling approximately $16,000 on two lawyers after discovering that they embedded hidden instructions within court filings intended to manipulate the court's artificial intelligence tools. The commands, written in white text on a white background, were invisible to the naked eye but were identified by court systems during processing.
The hidden prompts instructed the AI to process the documents only superficially and to ignore subsequent internal commands. While the attempt to manipulate the system did not ultimately affect the case's outcome—the plaintiff won by default after the defendant failed to respond—the court ruled the behavior a violation of judicial integrity. In addition to the fine, the lawyers were reported to the regional bar association and the labor court's internal affairs department.
This incident highlights rising concerns regarding the security of AI integration within the legal sector. Courts in Brazil have faced previous instances of users attempting to influence AI via prompt injection. These developments underscore the ongoing difficulty in balancing the efficiency of automated legal processing with the need to ensure that digital evidence and filings remain tamper-proof.