CPS cited a fake case in arguments to prosecute an 11-year-old

A citation that did not exist

Dhillon cited CPS – R v J [2024] EWCA Crim 3529. The case was entirely fictitious. In October, she acknowledged that she had cited the wrong authority and said she meant a real judgment from 15 years earlier.

The CPS decision to prosecute was made in April and confirmed by the Chief Crown Prosecutor. The boy’s lawyers challenged it, arguing that prosecutors had not taken account of submissions from social workers, youth workers and a court psychologist.

According to The Guardian, an AI-generated summary returned during an internet search supplied details of the invented case. The CPS later said the error arose during online legal research.

Brian Flanagan, an associate professor of law at Maynooth University in Ireland who studies judges’ use of AI, called it a “serious failure”. A fabricated source had been used to support the prosecution of a child, he said, raising questions about trust in a legal system that depends on authentic judgments. AI can save time, Flanagan added, but that also increases the risk of this kind of mistake. Errors by professionals are serious; errors by people working for the state carry particular weight because of the state’s power.

The High Court said the false citation did not change the outcome of the challenge. The boy’s case is due to be heard in a youth court at the end of this week.

A second CPS incident

This is the second known AI-related citation incident involving the CPS, the main public body responsible for criminal prosecutions in England and Wales. In February, the service acknowledged that one of its lawyers had cited nonexistent legal sources in an extradition case involving two Romanians convicted of minor offences. AI was considered a likely cause. The false cases were discovered after opposing lawyers could not find them; the CPS concluded the risk of a repeat was low.

The High Court had already warned about AI hallucinations in housing and banking proceedings. In June 2025, it urged the Bar Council and Law Society to consider urgent measures. Dame Victoria Sharp, then president of the King’s Bench Division, warned of “serious consequences for the administration of justice and public confidence in the justice system” if AI was misused. Lawyers could face penalties ranging from public reprimand to contempt proceedings or referral to the police, she said.

The CPS says its rules prohibit using AI to make legal decisions and require people to check the accuracy of its outputs. After discovering the citation, the service said it corrected the error, conducted an internal review to establish what happened and check whether other cases were affected, and reminded staff to cite judgments from reliable legal sources.

The gap between rules and practice

The incident brings the UK register of AI-related errors to 101 entries. Matthew Lee, a barrister at Doughty Street Chambers who maintains the register, includes confirmed, suspected and disputed errors, as well as false legal citations linked to them. He believes the recorded cases may be only “the tip of the iceberg”, since many lower-court decisions are never published.

I think the central weakness here is not that a search tool produced a false case; it is that a citation reached formal prosecution arguments without being checked against a reliable source. The CPS says its safeguards require human verification, but the account of this incident does not explain where that check failed. That gap matters more than the promise of low recurrence: the decision concerned prosecuting a child, and the authority used to support it was invented.