[Plaintiff Embeds Invisible AI Instructions in Court Documents to Manipulate Automated Review, Loses Electronic Filing Privileges] Deep Tide TechFlow reports that on August 15, according to The Decoder, a self-represented plaintiff in Connecticut, Matthew Elliott, embedded invisible AI instructions in court documents in an attempt to influence potential AI-based automated review systems. These instructions were presented in white font size 3 on a white background, making them nearly imperceptible to the human eye but fully readable by AI. The content was designed to guide the AI to produce outputs aligned with his claims and overturn a prior clerk's dismissal decision. The court discovered the manipulation due to unusual blank spaces in the document, which led to the exposure of this behavior. Judge Walter Spader Jr. likened this act to "secretly communicating with the jury" and noted that Connecticut courts do not actually use AI to review documents, so the attempt had no impact on the outcome. However, the intent to manipulate itself was deemed problematic. As a sanction, the court revoked Elliott's electronic filing privileges, requiring all his submissions to be made in person and on paper.