Litigant hides prompt injection in filing
Archive item — written before sources were shown.
A Connecticut judge caught a plaintiff hiding AI instructions in filings meant to sway any AI reviewing the case, the first known attempt of its kind.
A Connecticut judge disclosed what he called the first known case of a US litigant hiding AI prompt injection in court filings. Plaintiff Matthew Elliott formatted text to be invisible to a human reader but legible to any AI system, instructing it to favor his arguments and disregard prior rulings against him. Judge Walter Spader Jr. said the tactic had no real effect, since Connecticut’s courts do not use AI to review filings, but sanctioned Elliott anyway after he kept hiding text, including jokes, in later filings even after being warned.
- 01Suspecting court of using AI, man injected prompts in filings to try to win casearstechnica.com · primary, citing Judge Spader's order
