Policy & Regulation

CPS Prosecutor Cited Imaginary Case in High Court After AI Search

A senior CPS prosecutor cited a fabricated case while arguing to prosecute an 11-year-old; the AI-generated citation marks the agency's second such error and the UK's 101st tracked incident.

‘A serious failure’: CPS prosecutor cites imaginary case to high court after AI search
‘A serious failure’: CPS prosecutor cites imaginary case to high court after AI searchCole Miller / unsplash
By Rebecca Stone6 min read

Why it matters

  • CPS prosecutor Ravinder Dhillon cited nonexistent case CPS – R v J [2024] EWCA Crim 3529 in the high court.
  • The fabricated citation originated from an AI summary delivered in an internet search.
  • The incident is the second known AI citation error by the CPS and brings a UK tracker to 101 entries.
  • In June 2025 the high court urged the Bar Council and Law Society to act on AI misuse as a matter of urgency.

A senior crown prosecutor cited a completely fabricated legal precedent before the high court while arguing that an 11-year-old boy should be prosecuted for the rape of his seven-year-old sister, court documents have revealed. The reference, "CPS – R v J [2024] EWCA Crim 3529," does not exist, and The Guardian understands the fictional case details came from an AI summary generated during an internet search.

The prosecutor, Ravinder Dhillon, works in the Crown Prosecution Service's rape and serious sexual offences unit. Her explanation for how the nonexistent case entered her submissions has been called "unsatisfactory" by two senior high court judges, and the episode has been described as "a serious failure" by an expert in AI and the law.

The case is only the second known instance of an AI-related citation error involving the CPS, the main public agency bringing criminal prosecutions in England and Wales. It brings to 101 the number of entries in a UK tracker of AI-related court errors maintained by Matthew Lee, a barrister at Doughty Street Chambers.

What happened in the case?

Lawyers for the boy, who was 11 at the time of the alleged rape of his younger sister, were challenging the CPS's decision to prosecute him. They alleged prosecutors had failed to take into account representations from social workers, youth workers and a forensic psychologist.

Dhillon argued in favour of the prosecution. She submitted a detailed reference to a case she called CPS – R v J [2024] EWCA Crim 3529. The case turned out not to exist.

The decision to proceed with the prosecution was taken by Dhillon in April and confirmed by the chief crown prosecutor. In October, Dhillon admitted the nonexistent citation had been made in error. She said she had in fact been referring to a completely different decision — a real one — made 15 years earlier.

The high court found that the CPS's reference to the fictitious case did not affect the outcome of the boy's challenge to the decision to continue with his prosecution. The case is listed to go to trial in the youth courts later this week.

How serious is the failure?

"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," said Brian Flanagan, associate professor of law at Maynooth University in Ireland, who investigates the use of AI by judges.

"It calls into question norms of trust in the law where we treat authorities as precious things. You can't make them up; it's out of bounds," Flanagan said. "AI increases the risk that is going to happen because of the time you can save. It is not great for any professional to do this, but when the professional is working for the state it takes on a different aspect, because the power of the state is immense."

The stakes are structural, not merely personal. Court citations function as a shared currency of trust: lawyers and judges rely on the existence and accuracy of cited authorities to build and test legal arguments. When a state prosecutor submits a fabricated case in a child rape prosecution, that trust operates under the authority of the state itself.

Is this the first time for the CPS?

No. The CPS admitted in February that one of its lawyers, acting in an extradition case involving two Romanians convicted of minor offences, cited nonexistent legal authorities, likely caused by AI. Those made-up cases were spotted only when opposition lawyers could find no trace of them. The CPS concluded after that incident that "the risk of recurrence was low."

Beyond the CPS, the latest case adds to dozens of others this year in which AI-related errors have plagued court proceedings. Most involve people who are not legally qualified and are representing themselves, often in employment tribunal cases.

Lee's tracker now lists 101 entries covering confirmed, suspected and disputed AI-related errors, as well as related false citations. He said the identified cases may be the "tip of the iceberg," because many lower court decisions are never published.

Lee also raised two further concerns. First, privacy and confidentiality risks arise from the CPS using AI, given the highly sensitive information the agency handles. Second, AI is creating additional work for barristers, who must now spend time checking the authenticity of legal authorities appearing in case papers where AI may have been used in their preparation.

What has the CPS said?

Asked for comment, the CPS said it maintains a strict AI policy "which limits its use so that it is not used to make legal decisions and there must be human checks for accuracy."

A spokesperson said the CPS "identified an error had occurred from conducting legal research using the internet."

"As soon as this error was identified, it was remedied and an internal review was carried out to establish how this had occurred and to check that no other cases have been affected," the spokesperson said. "Following the rapid review we shared learnings, including that case law must be cited from trusted legal sources."

It is understood the internet search delivered an AI summary that contained the fictional case details.

What are the courts doing about AI hallucinations?

In June 2025, the high court responded to AI hallucinations in cases spanning housing to banking by calling on the Bar Council and the Law Society to consider steps to curb the problem "as a matter of urgency."

Dame Victoria Sharp, then president of the King's bench division, warned of "serious implications for the administration of justice and public confidence in the justice system if artificial intelligence is misused." She said lawyers misusing AI could face sanctions ranging from public admonishment to contempt of court proceedings and referral to the police.

Why does this case matter beyond the CPS?

The incident moves the AI hallucination problem one step up the institutional ladder. Until now, most documented UK cases involved litigants in person — people without legal training representing themselves. This case involves a senior prosecutor in a specialist sexual offences unit, acting for the state in a prosecution of a child, in the high court.

The CPS's own answer — that case law must be cited from trusted legal sources and that human checks are required — now stands against a second documented failure in under a year, in an agency that had previously assessed the risk of recurrence as low. With 101 tracked cases, a trial scheduled in the youth courts this week, and the Bar Council and Law Society still weighing the judiciary's June 2025 call for urgent action, the pressure for binding verification rules in UK court practice is likely to grow.

By Rebecca Stone, AI In Context

Original: bailii.org

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

391 articles

Related articles

  1. England's Top Civil Judge Says AI Should Decide Simple Court Cases
  2. Aachen court makes it two: German judges keep holding Google liable for AI answers
  3. Co-op Legal Services rates staff calls with AI scores
  4. AI Labs Sell Legal Tools as Courts Keep Punishing Fake Citations
  5. WA police scanned 1.38 million faces, made 122 arrests and logged nine false alerts

« Previous article