CPS prosecutor cited a fictional case after AI search in child prosecution dispute

The Crown Prosecution Service requires human checks on AI accuracy. Yet a senior prosecutor cited a fictional case while defending the decision to prosecute a child, the Guardian reported on November 3. The citation came from an AI-generated internet search summary, according to the newspaper; the High Court said it did not affect the challenge’s outcome.

A nonexistent authority in a child’s case

Ravinder Dhillon, a senior crown prosecutor in the CPS rape and serious sexual offences unit, submitted the reference CPS – R v J [2024] EWCA Crim 3529. Despite its detailed legal citation, the case did not exist.

The proceedings concerned whether the CPS should prosecute a boy who was 11 at the time of the alleged rape of his seven-year-old sister. His lawyers challenged that decision, alleging prosecutors had failed to consider representations from social workers, youth workers and a forensic psychologist.

Dhillon had decided to proceed in April, and the chief crown prosecutor confirmed that decision. In October, she acknowledged the citation was an error. She said she had intended to refer to a different, genuine decision made 15 years earlier. Two senior High Court judges described her explanation as “unsatisfactory.”

The High Court said the fictitious reference did not affect the outcome of the boy’s challenge to continuing the prosecution. His trial in the youth courts was listed for later that week, the Guardian reported. The challenge concerned the decision to prosecute, rather than a determination of guilt.

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.

Brian Flanagan, associate professor of law at Maynooth University, speaking to the Guardian

Research safeguards, not just decision rules

The CPS said the error arose during legal research using the internet. The more specific account—that a search delivered an AI summary containing the invented case details—comes from the Guardian’s reporting. The incident concerns material used to support a legal argument, not an admission that AI made the prosecution decision.

Responding to criticism, the service said its policy limits AI use so that it does not make legal decisions. It also requires human accuracy checks. Those safeguards cover different risks: who makes the decision, and whether the information used in legal work is correct.

The CPS said it corrected the error as soon as it was identified and conducted an internal review. That review sought to establish how the mistake happened and check whether other cases had been affected. The service said it then shared lessons with staff, including that case law must come from trusted legal sources.

An earlier error, and the cost of checking

This was the second known CPS incident involving nonexistent legal authorities, according to the Guardian. In February, the service admitted that a lawyer in an extradition case had cited invented authorities, likely because of AI. Opposing lawyers discovered the problem when they could not find the cases. The CPS then concluded that the risk of recurrence was low.

Barrister Matthew Lee’s UK tracker reached 101 entries with the latest incident. That is not a count of proven AI fabrications: it includes confirmed, suspected and disputed AI-related errors, alongside related false citations. Lee warned that unpublished lower-court decisions could leave other incidents unseen. He also said AI creates extra work for barristers who must check whether authorities in case papers are genuine.