When a new artificial intelligence tool is developed for the legal sector, its relevance is not measured solely by its technical performance. It also depends on its ability to meet the needs of the professionals who will use it in their day-to-day practice. Deliverable D4.1 – Chatbot Presentations Comprehensive Report provides particularly interesting insights in this regard by analysing the feedback gathered during the EuroLegalBot presentations organised in Italy, France, and Greece.
Beyond their role in raising awareness of the project, these events provided a valuable opportunity to observe the expectations expressed by judges, prosecutors, lawyers, academics and other stakeholders in European judicial cooperation. A cross-cutting analysis of the discussions reveals several key insights that extend well beyond the scope of the EuroLegalBot project alone.
Remarkable convergence despite differing national approaches
Whilst the presentations were tailored to national contexts, they centred on a coherent set of themes: the value of artificial intelligence-based tools for the European arrest warrant, the need for legally reliable information, the importance of validated and verifiable sources, multilingual interaction, assistance with legal research, and the maintenance of human oversight in sensitive criminal proceedings.
The differences in approach between the three countries are themselves instructive.
In Italy, the chatbot was presented both as an operational tool and as a training tool. Emphasis was placed on the use of verified legal sources, text- and voice-based interactions, multilingual capabilities, the export of conversations and the generation of structured legal documents, particularly in the field of the European Arrest Warrant.
In France, the presentation placed greater emphasis on the practical aspects of the tool for judges, prosecutors, and legal practitioners dealing with cross-border judicial cooperation. The aim was also to familiarise future participants in the pilot scheme with the methods for evaluating the chatbot.
In Greece, the approach was more technical and legal in nature. The presentations covered the system’s architecture, including Retrieval-Augmented Generation (RAG) mechanisms, natural language processing (NLP), legal semantic search, the traceability of responses, the system’s auditability, as well as issues relating to compliance with the GDPR, the AI Act and European digital sovereignty.
What practitioners consider essential
One of the key contributions of the Presentation Content Analysis is to identify the criteria that appear to be decisive for the use of a legal assistant in the context of European judicial cooperation.
The first relates to the very nature of the role assigned to the chatbot. In all three countries, EuroLegalBot is viewed as a support tool designed to assist with legal research, guidance, training and procedural preparation.
The second key finding relates to the trust placed in the information provided. Discussions show that practitioners attach particular importance to the inclusion of validated legal sources and the traceability of the references used. For future users, the quality of a response is measured not only by its apparent relevance, but also by the ability to identify its legal basis.
The multilingual aspect emerges as a third key element. In the context of the European Arrest Warrant, the ability to interact with the tool in several languages and to access information from different legal systems is seen as a particularly useful factor for cross-border judicial cooperation.
Finally, participants emphasised the need for responses that are contextualised, up-to-date and tailored to the relevant jurisdictions. The report thus highlights that the outputs of a legal assistant must be sensitive to the applicable legal context and take account of the diversity of national systems.
Maintaining human oversight as a guiding principle
Among all the topics analysed, one element emerges consistently across the three countries: artificial intelligence must support legal work without replacing professional judgement.
The summary provided in the report thus indicates that participants consider that AI should assist with research, procedural preparation and access to information, whilst leaving practitioners responsible for legal analysis and institutional decision-making.
This conclusion is likely the main cross-cutting lesson of the deliverable. It serves as a reminder that the value of a tool such as EuroLegalBot lies less in the potential automation of decision-making than in its ability to facilitate access to reliable, verifiable and actionable legal information for legal professionals.
Deliverable D4.1 shows that the national presentations of EuroLegalBot were much more than a mere communication exercise. Thanks to the structured analysis of the discussions, they made it possible to identify the conditions under which an AI-based legal assistant can be perceived as useful in the context of European judicial cooperation.
The lessons learnt from the events organised in Italy, France, and Greece converge around five requirements: a tool to support research and procedural preparation; verified and traceable sources; multilingual capabilities suited to cross-border cooperation; contextualised and up-to-date responses; and the maintenance of human oversight over legal analyses and decisions. These elements provide valuable guidance for the next stages in the development of EuroLegalBot and, more broadly, for the design of digital tools intended for practitioners in the European legal system.
Français