tl;dr: A Brazilian labor court fined two law firms roughly $16,000 (10% of the claim's value) for embedding a hidden prompt-injection attack in a court filing—white text on a white background instructing any reviewing AI to respond superficially and not challenge the documents. The court's Galileu AI system detected the hidden instruction, flagged it, and blocked processing. The ruling, believed to be the first of its kind globally, grounded the sanction in procedural good-faith obligations under Brazil's Civil Procedure Code.