How a UK rape trial sparked a warning over AI-generated testimony

• A UK rape trial collapsed after the complainant admitted using an AI chatbot to prepare for cross-examination.
• The Court of Appeal said AI-assisted preparation could amount to "witness coaching", which is banned in UK courts.
• Judges warned that AI could unintentionally reshape an honest witness's memory or help a dishonest witness improve their testimony.

As artificial intelligence becomes part of everyday life, UK courts are beginning to confront a new question: Can an AI chatbot influence a witness's evidence?

The issue came into focus after the Court of Appeal warned that witnesses should not use AI tools such as ChatGPT, Claude or Gemini to prepare for giving evidence. The court said such use could amount to "witness coaching", a practice that is strictly prohibited because it risks influencing testimony.

The warning followed a rape trial at Lewes Crown Court, where the complainant, identified only as C, admitted using an AI chatbot to prepare for cross-examination. The defendant was referred to as R.

When AI crosses the line

During the trial, C sought to submit screenshots from her phone as evidence. That led to the disclosure of chat logs showing she had used an AI chatbot to prepare for questions under headings including "On what happened that night", "On Consent and Capacity", "On Physical Evidence" and "On Your Behaviour After", according to court documents.

Lawyers for the defence argued that it was no longer possible to know how much of the witness's evidence reflected her own recollection and how much may have been influenced by the chatbot. The Crown Court accepted that concern, bringing the trial to an end.

Although the Court of Appeal later ruled that the case should be retried at the earliest opportunity, it agreed that the wider issue raised by the use of AI required careful attention.

In its judgment, the court reportedly said that all witnesses, whether for the prosecution or the defence, should be firmly discouraged from using AI to prepare their evidence.

The judges also warned that an honest witness could unintentionally change the emphasis of their evidence after reading AI-generated responses or begin to believe events happened differently. A dishonest witness, the court added, could use AI to quickly identify ways to make their testimony appear stronger.

A warning for the AI era

Witness coaching has long been prohibited in the UK because evidence is expected to come from a witness's own memory rather than a rehearsed or influenced version of events.

Lawyers are allowed to explain court procedures and conduct mock cross-examinations, but they cannot shape or rehearse a witness's evidence. The principle has existed for years, but the rise of generative AI is creating a new challenge for courts because chatbots can suggest wording, identify inconsistencies and generate answers in seconds.

The ruling is one of the clearest judicial warnings yet about the risks of relying on AI during legal proceedings.

As tools such as ChatGPT, Claude and Gemini become increasingly common in workplaces, schools and everyday life, the judgment suggests that courts may treat AI differently from other study or preparation aids. Rather than simply helping people organise information, judges appear concerned that AI could alter how witnesses remember, explain or present their evidence, raising fresh questions about the role of generative AI in the justice system.