Abstract

Artificial Intelligence (AI) applied to the legal domain is gaining ground. AI is argued to be particularly helpful with labour-intensive activities and repetitive tasks. Amongst the various AI solutions, ChatGPT has gathered momentum and its acclaimed advantages are, amongst others, document generation and contract review. This paper wishes to assess the effectiveness of two chatbots in contract drafting. To this aim, ChatGPT (by OpenAI) and Gemini (by Google) are prompted to write two supply contracts each, the first one written in English according to the laws of England and Wales, and the other one still written in English, but according to the laws of the Republic of Italy. The paper finds that each contract pair generated by the AI solutions is identical, thus disregarding legal-system specificities and language conventions. Therefore, the drafts present a number of fallacies, either because they are incomplete, or because they include clauses that are redundant or inapplicable in a given legal system. The paper claims that human intervention and supervision are necessary to tackle the shortcomings generated by AI, at least as far as contract generation is concerned and with regard to the two chatbots addressed.