A self-represented plaintiff in Connecticut, Matthew Elliott, embedded invisible AI instructions in official court filings to sway a potential automated review in his favor. The hidden text, formatted in 3-point white text on a white background, was designed to direct a hypothetical AI system to align its output with the submitted filing and correct a prior clerk's denial. The court discovered the manipulation due to unusual whitespace in the documents. Source: thedecoder
The court's discovery of the hidden text led to a hearing, where Judge Walter Spader Jr. warned Elliott against hiding text in court documents. Elliott continued to include hidden messages, including a YouTube link and mocking comments, in later filings. He described the original move as an 'audit' of a possible AI review system and the later messages as 'invisible jokes.' The judge emphasized that such covert communication is akin to secretly influencing a juror through an automated agent. Source: thedecoder
Connecticut courts do not currently use AI systems to review or rule on filings, so the hidden instructions had no impact on the case's outcome. However, the judge highlighted the ethical concerns of attempting to manipulate AI tools, noting that language models can reinforce flawed legal reasoning. The court revoked Elliott's electronic filing privileges, requiring him to submit all documents in person. Source: thedecoder
