Judges in England and Wales have been given the green light to use generative AI tools like ChatGPT for administrative tasks but warned against using them for conducting legal analysis.
The guidance, issued by the judiciary on Tuesday, advised judges to check any information provided by an AI tool for accuracy before it is relied upon.
Generative AI tools are known to provide incorrect information, including in the legal sector.
Two New York lawyers who submitted a legal brief containing fake case citations generated by ChatGPT were sanctioned over the summer.
The guidance warned judges against entering private or confidential information into a chatbot and said that AI tools are a “poor way of conducting research” to find new information that the person is not already familiar with.
However, the official guidance recommends using AI tools for summarising large bodies of text, writing presentations and administrative tasks.
Master of the Rolls, Sir Geoffrey Vos, said the judiciary “must embrace the adoption of developing technologies in our justice system, whilst ensuring that AI is used safely and responsibly”.
He added: “Judges do not need to shun the careful use of AI. But they must ensure that they protect confidence and take full personal responsibility for everything they produce.”
Generative AI systems are trained on enormous datasets. They identify patterns and structures within data to provide information in response to prompts. Because they base responses on predictions, generative AI models sometimes provide false information known as ‘hallucinations’.
“For now, while AI offers practitioners a new tool with extraordinary potential, we are some way from being confident that AI-generated material can be deployed unsupervised,” said James Levy, partner at law firm Ashurst.
“That is to say at this stage of development, all outputs purporting to represent legal opinion need to be reviewed, and verified carefully by qualified and experienced practitioners.”
Startups including Robin AI have seen the opportunity in bringing generative AI into the legal sector.
“This guidance is the first step in a proposed suite of future work to support the judiciary in their interactions with AI and to explore how AI will impact on the legal system and our courts,” said LJ Birss, deputy head of civil justice.
“It will be reviewed and updated as we learn more and as the technology continues to develop.”