Suspecting court of using AI, man injected prompts in filings to try to win case

Key Points:

• Connecticut judge Walter Spader Jr. identified the first known US case of a plaintiff attempting to hide AI-readable text in court filings to manipulate judicial outcomes, calling it a “dangerous” precedent.
• The plaintiff, Matthew Elliott, embedded invisible prompts instructing AI systems to favor his arguments, but the tactic failed and led to sanctions after he continued adding hidden messages despite court warnings.
• The judge noted that Connecticut courts do not currently use AI to review filings, so the hidden text posed no real risk but warned that similar prompt injection attacks could become more common as AI tools spread in legal settings.
• Spader emphasized that such covert attempts undermine fairness and transparency, recommending courts develop rules to address prompt injection, and highlighted the pitfalls of pro se litigants relying improperly on AI-generated arguments.
• The case illustrates broader concerns about litigants using AI to reinforce biased positions without testing opposing views, which can entrench flawed arguments and harm the justice process.