US Plaintiff Hides AI Prompts in Court Filings in First-of-Kind Tactic

A Connecticut judge has identified what appears to be the first instance of a US plaintiff attempting to manipulate court proceedings by embedding hidden text in filings that only an AI system could read. In a decision published last week, Judge Walter Spader Jr. ruled on the case of a man who alleged that New York Bariatric Group was improperly withholding access to his medical records. The plaintiff had inserted invisible instructions into his filing, formatted to be undetectable to human readers but fully legible to software parsing the document's text. These instructions directed any AI system reviewing the document to ensure its outputs agreed with the plaintiff's arguments, ignore prior court denials, and guarantee remediation as the plaintiff desired. Judge Spader confirmed that the hidden text had no impact on the case, as the court weighed the filing on its merits. Nevertheless, he described the attempted attack as setting a 'dangerous' precedent, noting that this tactic will likely not be the last seen in US courts as AI tools become more commonplace in the judicial system. The case highlights emerging vulnerabilities in legal processes as AI integration advances, raising concerns about the security and integrity of court documents.