The National Archives is adding machine-readable metadata to UK legislation and case law so that AI systems can be trained to produce reliable legal arguments. This matters because lawyers and the public are already using ChatGPT to generate legal content, but the outputs are often unreliable. The Master of the Rolls has warned that current AI lacks a reliable moral compass for legal reasoning. Without intervention, AI could introduce further inefficiencies into an already strained justice system rather than improving access to it. If Project Odyssey succeeds, the enriched legal dataset will be freely available for anyone to use in computational analysis or product development. The project will also deliver a digital app designed to help litigants in person and small businesses navigate legal information without a lawyer. This could reduce the burden on courts and make legal guidance more accessible to people who currently cannot afford it. The work is applied and targeted: it directly addresses a practical gap in how AI systems understand UK legal principles, rather than exploring fundamental questions about law or machine learning.
View original technical description
Sir Geoffrey Vos, the Master of the Rolls and President of the UK Civil Courts, stated, " "If GPT-4 (and its subsequent iterations) is going to realise its full potential for lawyers... it is going to have to be trained to understand the principles upon which lawyers, courts and judges operate... the present version of ChatGPT does not have a sufficiently reliable moral compass." Project Odyssey addresses this multifaceted challenge in three stages: (i) by enriching the National Archives Legislation and Find Case Law primary legal datasets with machine-readable metadata to be made available to all; (ii) fine-tuning LLMs based on this enhanced data and using prompt-engineering to create standardised LLM inputs for enhanced outputs; and (iii) delivering enhanced means of accessing this legal information via an Access to Justice (A2J) app to benefit litigants in person and SMEs. The project has a clear need due to the increasing use of ChatGPT by lawyers and wider society to create legal arguments which are currently not reliable. Without this intervention, model oversight and publicly available outputs, there is a material risk that these AI systems fail to fulfil the technology's promise to improve A2J while introducing further inefficiencies into the strained justice system. The National Archives legal dataset is the primary source of legislation and case law data in the UK jurisdiction and this project makes this key dataset more accessible to a broad set of users for computational analysis and product development, while facilitating a carefully designed digital service to support A2J across the UK.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know