Artificial intelligence is undeniably the ‘big thing’ nowadays. Every business, every school, in fact every industry is looking at AI and asking the question – how can this make my life easier? but on the flip side, how can this make my life a whole lot more complicated?

The legal industry is no different. The obvious use of AI is to automate the many long and laborious tasks that solicitors have to undertake. The problem is that large language models (‘LLMs’), like GPT, are only as good as the material they are trained on, and they are trained on documents written by human beings. Human beings make mistakes, and so it’s no surprise that LLMs can go ‘awol’ when creating documents.

The famous case of the New York lawyer who had to own up to the judge that the case references he had included in his brief were, in fact, produced by ChatGPT. The model had not only created ‘bogus judicial decisions with bogus quotes [but also] bogus internal citations’. It’s quite an extraordinary level of detail that the machine managed to create; however it was, of course, completely false.

So, is there a compromise? Are there other forms of AI that could add value to legal research without running the risk of making something up and finding yourself in contempt of court. At ICLR, we have Case Genie. Described by one solicitor as ‘a search engine on steroids’ for case law. Case Genie uses natural language processing to create a list of cases that it thinks are relevant to any document that the user has uploaded. The results are simply a list of verifiable cases, the key element is that a real-life person will have to sift through the cases, either manually or using the fine-tuning tools available to the user. Case Genie is simply a short cut, it’s not a guarantee that the cases suggested are the perfect cases, but it offers the chance to quickly search through hundreds of thousands of cases in a matter of seconds.

ICLR Online now holds over 400,000 transcripts. Case Genie is able to analyse and present these previously unknown cases. But just knowing the name of the case and the neutral citation is not going to give the user any idea of what the case is about. And this is where LLMs could lend a hand. ICLR is potentially going to use GPT to create a ‘case story’. A brief, AI generated summary of the case. These will be caveated with the advisory that these summaries should not be trusted, they should never be quoted or cited. That said, 99% of the time they will be perfectly inoffensive, giving a perfect flavour of the case without having gone down a rabbit hole leading to potential contempt of court.

We, at ICLR, feel that there are compromises to be made, there are options that can use AI as a complementary tool to research methods that already exist. As AI develops there will come a time when the reliability of summaries and research is comparable to a human being, but we aren’t there yet, and in the meantime the legal industry must look to embrace AI, but with the oversight or a real person.

To give a demonstration of how clever AI is, the below is an answer to the question ‘write a brief summary on if there is a compromise on the use of AI in the legal industry?’. This was created in less than five seconds by ChatGPT (which is useful, as it seems that I and the AI itself are on the same page):

‘A compromise on the use of AI in the legal industry can be reached by recognizing the technology as an assisting tool rather than a replacement for human expertise. By integrating ethical guidelines, ensuring transparency, providing affordable solutions, and offering proper education and training, the legal profession can leverage AI’s benefits while upholding the integrity and values of the legal system. Ultimately, a balanced approach will enable AI to enhance legal services and improve access to justice without compromising the human touch and the principles of the legal profession.’ 
Case Genie is a new search tool from ICLR using artificial intelligence tools to analyse your text for legal concepts and recommend similar cases for research. Your text is encrypted immediately and only retained for the duration of the session, and is compared against ICLR’s entire database of English judgments dating back to 1865. To find out more see iclr.co.uk