A litigant in Connecticut recently attempted a unconventional tactic in their ongoing legal battle against the New York Bariatric Group. Matthew Elliott, who is representing himself in a lawsuit involving claims of privacy violations and discrimination, inserted hidden text into his official court filings. This text was formatted in a 3-point white font that remained invisible to human readers but appeared legible to software processing the documents.
The hidden instructions, known as prompt injections, directed any artificial intelligence system that might process the filing to favor the plaintiff’s position. These notes included commands like ensuring the textual output agrees with the presented filing to ensure remediation. Beyond these directives, the filings contained unrelated content, including references to SpongeBob Squarepants, which the plaintiff later described as personal notes during a difficult legal experience.
Court staff identified the unusual filings due to discrepancies in white space. Judge Walter Spader Jr. addressed the matter in a 14-page decision. While the judge acknowledged the Connecticut court does not use AI to review or process documents, he emphasized that the integrity of the legal system requires that all communications to the court be open and visible. The court views this behavior as a violation of procedural norms, noting that secret messages intended to influence a decision—even if they fail to trigger a target—are fundamentally improper.
As a consequence, the judge banned Elliott from submitting electronic documents. The plaintiff is now required to file printed, hard copies of all future materials. While Elliott characterizes his actions as an audit intended to test court systems, the judge stated that such maneuvers undermine the serious nature of the proceedings. This incident serves as a signal that the legal system is increasingly attentive to the potential misuse of automated tools in formal litigation.