I suppose it was only a matter of time.

Here is a short post by Matthew Lee – a barrister who is tracking all things AI in law so you don’t have to – about how chat GPT prompts were adduced in evidence in family proceedings, much in the same way as internet search history is often relied upon.

Matthew’s post is here, and the original judgment he is writing about is here.

As Matthew points out, Chat GPT prompts are not quite the same as internet searches, and their meaning and what they might reveal about a person’s motivations will be very fact specific – but it seems to me that, as with internet searches, they do have potential in some cases to be really quite important evidence. The most obvious example is queries in the aftermath of an unexplained injury by a carer, which reveal their knowledge of injury or of particular mechanisms. The circumstances in this case were much more obviously susceptible to multiple different explanations, but that doesn’t mean that these searches will always be irrelevant.

Here the material was produced by (it appears) a party taking screenshots of prompt history without the user’s knowledge. I suppose we will now have to start thinking about whether our instructions to forensic experts tasked with forensic download work, will also have to include requests for retrieval of AI prompt histories, too. I have certainly seen increasing evidence that participants in family proceedings are using such products in connection with family proceedings, whether that is because their materials read like the words of a robot who has swallowed a legal dictionary, or because their phone downloads reveal screenshots of chat gpt results, so it is the logical next step.