CPS Prosecutor Cites Fictitious Case in High Court

A senior Crown Prosecution Service prosecutor cited a completely fictitious legal case during a High Court dispute over the prosecution of a child accused of raping his younger sister, prompting renewed concern about the growing risks of artificial intelligence errors in legal proceedings.

The nonexistent precedent was presented as part of the prosecution’s argument in a highly sensitive case involving an 11-year-old boy who was accused of raping his seven-year-old sister. Court documents later established that the legal authority cited by the prosecutor did not exist. The incident has since been described by an academic specialising in law and artificial intelligence as a “serious failure”, while two senior High Court judges criticised the explanation given for how the false citation entered the case.

Ravinder Dhillon, a senior crown prosecutor working within the CPS rape and serious sexual offences unit, had been arguing in support of continuing the prosecution of the child. During proceedings, she referred to a case identified as CPS – R v J [2024] EWCA Crim 3529. The citation appeared to present the matter as an authoritative Court of Appeal criminal decision, but no such case could be found.

The issue emerged after lawyers representing the boy challenged the CPS decision to prosecute him. They argued that prosecutors had not properly considered representations from social workers, youth workers and a forensic psychologist before deciding that criminal proceedings should continue.

The prosecution decision had initially been made by Dhillon in April and was subsequently confirmed by the chief crown prosecutor. In October, however, Dhillon acknowledged that the citation was an error. She explained that she had actually intended to refer to a different judgment, one that was genuine and had been issued about 15 years earlier.

The circumstances surrounding the false citation have heightened concern because the reference was used in an argument supporting the prosecution of a child. Although the High Court ultimately concluded that the fictitious authority had not affected the outcome of the boy’s challenge, the incident has raised broader questions about the standards expected when public prosecutors rely on legal research.

Brian Flanagan, an associate professor of law at Maynooth University in Ireland who researches the use of artificial intelligence by judges, described the episode as particularly serious because the fabricated authority had been used as a legal source in a case involving the state’s power to prosecute a child.

Flanagan said legal authorities occupy a special position in the justice system and cannot simply be invented or presented without verification. He also warned that artificial intelligence could make such failures more likely because of the speed with which AI-generated material can be produced during legal research.

The concern is not simply that an incorrect reference appeared in a court document. Legal proceedings depend heavily on the accuracy and traceability of judicial decisions, statutes and other authorities. Lawyers, judges and opposing parties rely on those materials to establish how previous courts have interpreted the law. A fabricated case can therefore undermine confidence in the legal reasoning being presented, even where the error does not ultimately change the result.

The latest incident is also significant because the Crown Prosecution Service is the principal public body responsible for criminal prosecutions in England and Wales. Mistakes made by members of the public representing themselves have accounted for many of the AI-related problems identified in British courts, particularly in employment tribunal cases. An error by a professional prosecutor acting on behalf of the state raises a different level of concern because of the authority and resources attached to the prosecution service.

The CPS has already faced another case involving apparently fabricated legal authorities. In February, the organisation acknowledged that one of its lawyers involved in an extradition case concerning two Romanian nationals convicted of minor offences had cited legal authorities that did not exist. The organisation concluded at the time that the problem was likely connected to artificial intelligence and assessed the risk of recurrence as low.

The new case suggests that concerns surrounding AI-assisted legal research have not disappeared. It has also raised questions about whether existing safeguards are sufficient when lawyers use internet search tools that may incorporate AI-generated summaries.

The CPS said it maintains a strict policy governing the use of artificial intelligence. Under that policy, AI cannot be used to make legal decisions, while human checks are required to establish the accuracy of information.

A CPS spokesperson said the organisation had identified the problem as an error arising from legal research conducted through the internet. Once the mistake was discovered, the reference was corrected and an internal review was carried out to determine how the error occurred and whether other cases might have been affected.

The organisation said lessons from that review had been shared internally, including the requirement that case law should be cited from trusted legal sources. The statement reflects a central concern surrounding AI-assisted research: even when technology is used only to locate information rather than make a legal judgment, material generated or summarised by automated systems still requires careful verification before it is relied upon in court.

The number of reported incidents involving AI-related legal errors in the UK has also continued to rise. A tracker maintained by Matthew Lee, a barrister at Doughty Street Chambers, now contains 101 entries involving confirmed, suspected or disputed AI-related mistakes and false legal citations.

Lee has warned that the recorded incidents could represent only a small proportion of the actual problem. Many lower court decisions are never published, meaning errors involving fabricated authorities may remain undiscovered unless another lawyer, judge or party notices that the cited case cannot be located.

The consequences extend beyond the immediate accuracy of court documents. Lawyers representing opposing parties may have to spend additional time checking every authority contained in material where AI-assisted research is suspected. That creates an additional burden on the justice system, particularly when legal professionals must investigate whether a cited case exists before they can engage with the argument it is being used to support.

There are also concerns about confidentiality. The CPS handles extremely sensitive information involving victims, suspects, children and witnesses. The use of AI systems in legal work therefore raises questions not only about inaccurate information but also about how confidential material is handled and protected.

The wider legal profession has already been warned about the risks. In June 2025, the High Court responded to a series of cases involving AI-generated hallucinations in areas ranging from housing to banking by urging the Bar Council and the Law Society to consider urgent measures to tackle the problem.

Dame Victoria Sharp, who was then president of the King’s Bench Division, warned that misuse of artificial intelligence could have serious consequences for the administration of justice and public confidence in the legal system. Lawyers who misuse AI can face professional or legal consequences, including public criticism, contempt proceedings and, in serious circumstances, referral to the police.

Against that background, the latest CPS incident is likely to add pressure on legal organisations to strengthen safeguards around AI-assisted research. The central issue is not whether lawyers can use modern technology, but whether every piece of legal information produced or surfaced by such systems is independently checked before it enters formal court proceedings.

For the boy at the centre of the case, the High Court has determined that the fictitious citation did not affect the outcome of his challenge to the decision to continue with the prosecution. His case is due to proceed to trial in the youth courts later this week.

The incident nevertheless leaves a wider question for the justice system: how can courts take advantage of rapidly developing artificial intelligence tools without allowing speed and convenience to weaken the reliability of legal authorities?

For prosecutors, barristers and judges, the answer increasingly appears to depend on maintaining a clear boundary between technology-assisted research and verified legal judgment. A search result may be generated within seconds, but the responsibility for putting an accurate authority before a court remains with the human professional who relies upon it.