Fabricated content produced by artificial intelligence tools has now found its way into the justice system. An experienced prosecutor at the Crown Prosecution Service (CPS), which operates in England and Wales, cited a court ruling that never actually existed as evidence in a sensitive case heard at the High Court.

When the cited ruling turned out to be imaginary, it caused significant alarm in legal circles. The incident has been recorded as one of the most concrete examples of how serious the consequences of AI errors can be.

Case for trying the child rested on a fictional ruling

gorsel_20261103_101036

At the centre of the case is a boy accused of sexually assaulting his 7-year-old sister. The boy was 11 years old at the time of the alleged offence, and the High Court was asked to consider whether he could stand trial. The defence argued that assessments by social workers, youth workers and forensic psychologists had not been taken into account by the prosecution, and sought to have the case dropped.

In response, senior prosecutor Ravinder Dhillon of the CPS's Rape and Serious Sexual Offences unit argued that proceedings should continue, citing a precedent labelled "CPS – R v J [2024] EWCA Crim 3529". However, an investigation revealed that no such ruling exists anywhere in the court records.

The prosecutor admitted the error in October. She said she had come across an AI-generated summary during an internet search and took the fictional ruling from it as a source, when what she had intended to reference was a real case from 15 years earlier. High Court judges were not persuaded by that explanation.

"A lamentable failure"

gorsel_20261103_101052

Associate Professor Brian Flanagan of Maynooth University's School of Law, who researches the use of AI in judicial processes, described the reliance on a fabricated source to justify trying a child as "a lamentable failure". In his view, turning to AI for quick results deeply undermines trust in the justice system, and the fact that the error was made by a public prosecutor is especially dangerous given the power the state wields.

The court announced that the fictional citation did not change the course of the case, and that the boy's proceedings in the youth court would continue. The picture remains worrying nonetheless: with this incident, the number of confirmed, suspected or disputed AI errors tracked in the UK has reached 101.

CPS management said the organisation has strict rules on the use of AI and that the technology must never be used to make final legal decisions. A spokesperson said the problem stemmed from an internet search, that an internal review was launched once it was identified, that no other similar case was found, and that staff were reminded of the principle that precedents must only be verified from reliable sources. Experts, meanwhile, take the view that processing sensitive and confidential information from prosecution files with AI tools poses serious privacy risks.