High court error: prosecutor cites imaginary case after AI search

Ravinder Dhillon, representing the Crown Prosecution Service (CPS), mistakenly referenced a nonexistent case while arguing whether an 11-year-old boy should be prosecuted for the rape of his seven-year-old sister. Court documents revealed the error, with the cited case, CPS – R v J [2024] EWCA Crim 3529, proving to be entirely fictitious.

Dhillon attributed the error to an AI-generated response from an internet search. Two senior high court judges deemed her explanation “unsatisfactory,” and an AI and law expert labeled the incident a “serious failure.” This follows a February admission by the CPS that a lawyer in an extradition case involving two Romanians cited fabricated legal authorities, also likely due to AI.

The decision to prosecute the boy was made in April and confirmed by the chief crown prosecutor. Dhillon acknowledged the error in October, stating she had intended to reference a real case from 15 years prior.

“This is a serious failure because it involves using a fabricated legal source to argue for the prosecution of a child,” stated Brian Flanagan, associate professor of law at Maynooth University. “It undermines the trust we place in legal authorities, which cannot be invented.”

The high court determined that the fictitious citation did not impact the outcome of the boy’s challenge to the prosecution, which is scheduled for trial later this week.

In June 2025, the high court urged the Bar Council and the Law Society to address the growing issue of AI-related errors, warning of potential consequences for the justice system and public trust. Sanctions for misuse of AI by lawyers, ranging from admonishment to contempt of court proceedings, were suggested.

The CPS stated it has a policy restricting AI use to avoid legal decision-making and requiring human verification. A spokesperson acknowledged the error stemmed from internet-based legal research and that an internal review was conducted to prevent recurrence.

This incident contributes to a UK tracker maintained by barrister Matthew Lee, documenting 101 confirmed, suspected, and disputed AI-related errors. Lee cautioned that this represents only a fraction of the problem, as many lower court decisions remain unpublished, and highlighted privacy risks associated with the CPS’s use of AI, adding to the workload for barristers who must now verify legal authorities.