The ethical dimension of AI systems in legal jargon translation: a preliminary assessment

Abstract Despite increasing legal emphasis on accessible language in contractual information, individuals continue to encounter legal documents, such as terms and conditions, imprints, and privacy statements, that remain difficult to understand. The emergence of large language models (LLMs) offers a new opportunity to translate legal jargon into accessible language and to support ongoing efforts to provide legal documents in accessible and inclusive languages. However, the use of LLMs for this purpose raises distinct ethical concerns, including hallucinations, personalization risks, and jurisdictional bias. This paper presents a preliminary ethics framework for evaluating LLM-based legal translation tools, grounded in EU regulations and aligned with the EU High-Level Expert Group’s Ethics Guidelines for Trustworthy AI. Building on the EU-funded AI4Pol project, which aims to develop an AI-powered tool for translating legal jargon into accessible language, we examine the ethical risks of such systems, focusing on their intended purpose and challenges related to technical robustness, privacy, algorithmic bias, and explainability.