Is AI Really the Greatest Threat to Witness Evidence?
Artificial intelligence has become far more than a simple search tool. For many people, it now acts as a friend, consultant, teacher, and sounding board for everyday questions, professional tasks, and even legal research. Against this backdrop, cases involving AI often attract significant attention.
AI as a witness coach
R v FGD [2026] EWCA Crim 918 is no exception. In FGD, a witness used AI to prepare for cross-examination. The AI generated anticipated questions and suggested answers designed to help the witness present a consistent account of their evidence. When this came to light, the trial judge concluded that the exercise amounted to witness coaching and stayed the proceedings as an abuse of process.
The Court of Appeal agreed that the witness's interaction with AI constituted witness coaching. However, it disagreed that the proceedings had to be stayed. The court held that the ordinary trial process remained capable of addressing any prejudice arising from the witness's use of AI. In other words, the use of AI that amounted to witness coaching did not automatically make a fair trial impossible.
Lessons from FGD
It is important to note that the use of AI does not necessarily amount to witness coaching. Had the witness merely used AI to help them understand court procedure or what to expect when giving evidence, the outcome may have been different. Such use might fall within the concept of witness familiarisation, which is permissible under certain conditions. The Bar Council of England and Wales has published detailed guidance in this regard.
The difficulty in FGD was that the AI did more than familiarise the witness with the court process. It generated anticipated questions and proposed substantive answers. Once a witness begins asking AI how particular issues should be explained, how inconsistencies should be addressed, or how certain questions ought to be answered, the exercise moves beyond familiarisation and into coaching. At that point, there is a genuine risk that the evidence ceases to be entirely the witness's own.
Importantly, this concern is not unique to AI. The same risk arises whenever a witness receives substantive guidance about their evidence from another person.
FGD is therefore not only a case about the use of AI in court proceedings, but also serves as a reminder of the law's continuing commitment to preserving the authenticity and integrity of witness evidence.
The use of AI in witness evidence
The judgment also reflects wider developments in the justice system. The Court of Appeal referred to the Civil Justice Council's recent consultation on the use of AI in civil proceedings, particularly in relation to witness statements. Significantly, the discussion focused not on AI itself, but on preserving the authenticity, integrity, and personal recollection of witnesses.
From a practical perspective, the case provides an important lesson for legal practitioners. Witnesses should be encouraged to focus on their own recollection of events and should be warned that they are not permitted to seek substantive assistance from AI tools, or indeed from any other source, when preparing to give evidence. This is true regardless of the witness's good faith. The risk lies not only in deliberate fabrication, but also in the unintended distortion of memory through external influence.
Please get in touch with any of the authors of this blog if you would like further information.
All of our thought leadership content relating to AI can now be found in our AI Hub – see here.
