Suspecting court of using AI, man injected prompts in filings to try to win case
A Connecticut judge has sanctioned a plaintiff after he hid prompt-injection text in court filings, apparently hoping to influence any artificial intelligence system used to read or assess them. Judge Walter Spader Jr. said the tactic did not affect the case’s outcome, as the Connecticut Judicial Branch does not use AI to review or decide filings, but warned that it could set a dangerous precedent as AI becomes more common in courts.
Matthew Elliott used tiny white text on a white background to instruct any AI reader to support his arguments, disregard earlier court denials and grant his preferred remedy in a dispute over healthcare records. After being warned, he continued inserting hidden messages that he described as jokes; the judge found this “serious litigation abuse”, imposed modest sanctions and rejected Elliott’s claim that he was merely auditing suspected court AI use.
In short
- Plaintiff hid AI-targeted instructions in court documents.
- Connecticut courts do not use AI to decide filings.
- Judge imposed sanctions and warned of wider risks.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
This behaviour constitutes serious litigation abuse warranting sanction. The individual deliberately hid manipulative prompts within court filings using white-on-white text and persisted after explicit warning, demonstrating contempt for court authority. Courts must establish firm boundaries against such manipulation tactics as artificial intelligence becomes more prevalent in legal proceedings, protecting judicial integrity and public confidence.
The case against
Whilst the methods were problematic, no actual harm occurred since the court demonstrably does not use artificial intelligence. The individual's claim that he was auditing for undisclosed AI systems raises legitimate questions about institutional transparency, particularly as courts increasingly deploy technology without full public disclosure. The modest sanctions imposed and characterisation of the communications as jokes suggest this was not viewed as egregious; courts would be better served by being transparent about their AI use.
Explained simply
Prompt injection is text designed to manipulate an AI system by giving it instructions inside material the system is asked to read. Hidden text can be hard for a person to notice, while still being readable by software.
Courts use filings, such as claims and evidence, to consider disputes and make decisions. The case involved a disagreement over healthcare records, and Matthew Elliott was the plaintiff, the person who brought the claim; Judge Walter Spader Jr. handled it.
The issue matters as AI tools become more common in legal work, including tasks that support the handling of documents. Hidden instructions raise questions about how courts can protect fair procedures and assess material that may be read by both people and software.