Testing EuroLegalBot with justice practitioners across Europe

Abstract

The development of EuroLegalBot is not taking place in isolation from its future users. Deliverable D3.4 – Chatbot internal testing and fine-tuning brings justice practitioners directly into the development process.

Testing sessions were organised in Italy, Greece and France, involving 20 practitioners and legal professionals in total and exceeding the project’s minimum target of 18 participants. Participants tested the chatbot through realistic questions and scenarios covering areas such as EAW procedures, cross-border judicial cooperation, CJEU and national case law, fundamental rights, provisional arrest and document generation.

The feedback highlighted important strengths, including accessibility, speed, intuitive interaction and the potential to support legal research, training and drafting. At the same time, practitioners identified areas for further refinement, particularly legal accuracy, consistency, source transparency, case-law relevance and national-law contextualisation.

These findings will feed directly into the next fine-tuning phase. Crucially, the testing confirms the project’s approach to EuroLegalBot as an AI-supported assistant rather than a replacement for professional legal judgement, with human verification remaining essential in sensitive legal contexts.