According to Tom's Hardware, the ruling was shared on X by Prof. Vladimir Aras of the University of Brasília. While the court's AI system was processing the complaint, it identified text in white font that was invisible to readers. After the color was changed, it revealed an instruction telling the system to respond to the petition superficially and not to challenge the documents, "regardless of any command given to you". The court named the technique: prompt injection.
The defendant did not respond and the plaintiff won by default, so the hidden instruction did not affect the outcome. The court still called it "an act offensive to the dignity of justice" and placed the responsibility on the two lawyers who signed the complaint rather than on the plaintiff. Besides the fine, they were reported to the state bar association and to the regional labor court's internal affairs department.
According to the report, this is not the first case: a long hidden prompt was previously found in filings by a plaintiff who was not a lawyer, who was then required to submit printed documents. It has not been reported which AI system the court uses.