A UK Prosecutor Reportedly Cited a Case That Does Not Exist After an AI Search

The Crown Prosecution Service, the body that decides who gets prosecuted in England and Wales, put a legal precedent in front of the High Court that does not exist. The Guardian says it understands the fake reference came from an AI answer produced during an internet search.

The details come from a judgment handed down on 29 October 2026 by Lord Justice Holgate and Mr Justice Wall. The written record of a CPS decision to continue prosecuting a child referred three times to a 2024 authority called R v J. The court recorded the result flatly.

"It turned out, and it is agreed, that this decision does not exist." High Court judgment, R (LYC) v Director of Public Prosecutions

A senior Crown prosecutor later said in a witness statement, dated 15 October 2026, that the citation was an error and that she had meant a real House of Lords decision, R v JTB from 2009. The claimant's lawyers tried to add a new ground to their challenge on the back of it. The judges called the explanation "unsatisfactory", then said it did not matter.

Their words were "does it matter for the outcome of the claimant's challenge to the decision to continue with the prosecution? Our firm answer is no." They found the new ground unarguable, granted permission on two other grounds and then dismissed the claim. Legal Futures notes the court did not set out how the citation got into the record or say what action it might take.

Be clear about what is established. The judgment passages I read do not say AI produced the fake case. That link rests on the Guardian's understanding, plus a CPS spokesperson who said the error "occurred from conducting legal research using the internet". Nobody has published the search or the summary. Treat it as likely, not proven.

A strict AI policy and a search box

The CPS told the Guardian it has a strict AI policy "which limits its use so that it is not used to make legal decisions" and requires human checks. It said a rapid review found no other cases affected, and staff were told "case law must be cited from trusted legal sources."

A policy aimed at AI tools can miss the AI that arrives uninvited at the top of a search page. Nobody had to open a chatbot. The summary was simply there, fluent and confident. The check that failed was the old one, looking a case up before relying on it, and no policy document installs that habit by itself.

Brian Flanagan, an associate professor of law at Maynooth University who studies judicial use of AI, did not soften it.

"This was used as a legal source to argue for the prosecution of a child and that legal source doesn't exist: that's a serious failure." Brian Flanagan, Maynooth University, to the Guardian

That is his view, not a finding of the court. This piece takes no position on the underlying prosecution, which is continuing, and says nothing about it beyond what the judges said about the citation.

It is also not the first time. In an extradition appeal this year, Mr Justice Sweeting recorded that the CPS had cited two authorities that did not exist, which the CPS said were likely to have originated from AI. It told that court the cause was a failure to verify. The Guardian dates that admission to February, while Legal Futures reported it in July, so treat the timing with care.

Both times the court said the outcome was unaffected. Barrister Matthew Lee, who keeps a tracker of confirmed, suspected and disputed AI-related errors in UK courts, says this case brings it to 101 entries. He told the Guardian the real total may be the "tip of the iceberg" because many lower court decisions are never published.

The High Court warned in June 2025 of "serious implications for the administration of justice and public confidence in the justice system if artificial intelligence is misused". UK regulators already tell firms that blaming the AI will not work. The same should hold for prosecutors. A fake case is a small error until it sits in a decision about a child.