Brazilian Court Sanctions Two Lawyers for Embedding Prompt Injection in Pleadings
A Brazilian labor court found that two lawyers deliberately concealed prompt injection instructions in pleadings and fined two law firms 10% of the amount in dispute, approximately $16,000. The case originated from a labor lawsuit at the 3rd Labor Court of Parauapebas, Pará. When a judge used Galileu, a generative AI tool promoted by the National Council of Justice, to review the pleadings, it identified hidden text in white font on a white background, whose content instructed the AI to respond superficially and not question the document. The court imposed penalties under Articles 5 and 77 of the Brazilian Civil Procedure Code, which set out the duties of good faith and not obstructing justice. After identifying the content, Galileu issued an alert and automatically blocked processing of the suspicious content.
Why read this
AI score 71
This is the first visible ruling by a court on prompt injection to manipulate judicial AI, presenting the court's basis for its findings and the scale of penalties.