Brazilian lawyers fined $16,000 for attempting to manipulate court AI via hidden prompts

A Brazilian labor court fined two lawyers approximately $16,000 for attempting to manipulate the court's AI tools. Professor Vladimir Aras of the University of Brasília shared the ruling, which found the lawsuit contained prompt injection instructions embedded in white-on-white text, invisible to the naked eye. This technique involved planting concealed commands to direct the AI to superficially process the complaint and disregard other instructions. Although the court deemed the act detrimental to judicial dignity, the injection had no effect as the defendant did not respond, leading to a default judgment for the plaintiff. Both lawyers have been reported to the state bar association and the regional labor court's internal department. This follows a previous instance where another plaintiff embedded a long hidden prompt, resulting in a lighter penalty and a requirement for future hard copy submissions. Separately, a state judge faced a federal lawsuit concerning AI-generated judgments, which the court dismissed citing judicial immunity.