ChatGPT and GPT-4: utilities in the legal sector, functioning, limitations and risks of foundational models

Artificial intelligence systems such as ChatGPT, the OpenAI chatbot, based on the language model family GPT (generative pre -trained transformers), as well as other solutions built on this technology and fine-tuned for specific tasks, have generated considerable interest across various sectors, including the legal sector. However, such models still feature important limitations and legal risks associated to their use, which must be considered in order to make a proper and legally responsible use of this technology. This work aims to familiarize the readers (men and women) with the configuration, architecture, and functioning of these systems, as well as their functionalities in the legal sector. It includes a review of their associated legal limitations and risks, with crucial practical implications in their application.