A judge in the US state of Connecticut announced the identification of the first instance of an attempt to trick artificial intelligence systems in court documents. In this case, the plaintiff had attempted to alter the course of the proceedings in their favor by inserting hidden text that is readable only by AI systems.
Walter Spader Jr., the judge in the case, announced that these hidden texts were formatted in a way that made them invisible to the human eye, but fully readable by text-processing software. These hidden instructions dictated the AI to align its outputs with the plaintiff's arguments and ignore the court's previous negative rulings. Although this action had no impact on the final ruling, the judge described it as a "dangerous precedent" that could threaten judicial security as AI tools spread in courts.