Digital transformation has reshaped many industries, but legal research has long remained labour-intensive and fragmented. Together with strong technological roots and a deep understanding of professional workflows, DeepLaw set out to change that.

How and when did the idea for DeepLaw arise?

The idea for DeepLaw arose from a concrete observation of everyday legal practice: legal research occupies a considerable proportion of the working time of lawyers and other legal professionals. Despite increasing digitalisation, research often remained time-consuming, fragmented and dependent on traditional keyword searches. At the same time, the volume of legal information continued to grow, while pressure for speed, efficiency and traceability intensified.

Against this backdrop, it became clear that conventional search solutions were no longer meeting the demands of modern legal work. The trigger for DeepLaw was therefore the clearly identified problem that legal professionals were required to devote valuable resources to research, even though the relevant information was fundamentally available in digital form.

In particular, the comprehensive and continuously updated integration of official legal sources such as Fedlex, Lexfind, as well as federal and cantonal court decisions, posed significant requirements in terms of ensuring data quality and guaranteeing a reliable foundation for structuring and synchronisation.

What vision did you have for DeepLaw at the beginning, and in what respects has this vision evolved over the course of development?

From the outset, the aim of DeepLaw was to make legal research in Swiss law noticeably simpler and faster. The product is specifically aimed at legal professionals such as lawyers, notaries and in-house legal departments, all of whom require precise and verifiable legal information on a daily basis.

The original vision was to create an innovative platform that allows users to pose complex legal questions in natural language and receive clear, well-founded answers from official sources within seconds. DeepLaw was not intended to provide opinions or simplified summaries, but to rely exclusively on original, directly verifiable legal sources.

Artificialy was also involved in the development of DeepLaw. Could you briefly explain who Artificialy are and what they contributed to the project?

Artificialy, headquartered in Lugano and Zurich, is a company founded in 2019 by Luca Gambardella and Marco Zaffalon, professors at IDSIA (Dalle Molle Institute for Artificial Intelligence) with decades of experience in artificial intelligence.

Artificialy is responsible of developing many AI components in the DeepCloud ecosystem and it also contributed to the development of DeepLaw by building the entire AI engine: specifically the AI model, the intelligent semantic search engine, the agentic pipeline, model optimization, and production deployment of the entire AI component.

What were the biggest professional or organisational challenges during development?

One of the greatest technical challenges was combining AI-driven technology with the stringent requirements of legal precision. Legal research demands reliability, transparency of sources and the ability to cite them – qualities that are often lacking in general AI or search solutions.

In particular, the comprehensive and continuously updated integration of official federal legislation via Fedlex, cantonal legislation via Lexfind, as well as the direct integration of decisions from the Federal Supreme Court and cantonal courts, placed high demands on ensuring data quality, structuring and synchronisation. At the same time, DeepLaw had to be developed to understand natural language, correctly process multilingual content and yet provide exclusively verifiable original sources.Englishcen

What role did feedback from users, clients or external partners play in the further development of DeepLaw?

Feedback from users and legal professionals played a central role in the ongoing development of DeepLaw. As the platform was designed specifically for legal practitioners, it was essential to understand working methods, expectations and concrete use cases from practice.

Feedback proved particularly valuable in optimising the user interface, refining the response logic and ensuring that the content provided was not only accurate but immediately usable in day-to-day legal work. This close alignment with the needs of the target group contributed significantly to establishing DeepLaw as a practical and efficient research solution.

What are you particularly proud of in the finished product?

DeepLaw fundamentally distinguishes itself through its consistent focus on Swiss law and exclusively on official, original legal sources. While other AI tools provide secondary opinions or unverifiable web content, DeepLaw offers directly citable statements together with access to the original sources at federal and cantonal level.

Particularly noteworthy is the combination of natural language processing, multilingual functionality including automatic translations, and a rigorous approach to data protection: queries are not used for AI training and remain entirely confidential on Swiss servers. This combination of efficiency, precision and security makes DeepLaw unique.

What happens next for DeepLaw following its launch? What next steps are already foreseeable?

Following the launch, the focus is on continuously developing DeepLaw and tailoring it even more closely to the needs of legal professionals. Foreseeable next steps include further optimisation of the AI-driven search and response logic, ongoing expansion and updating of legal sources, as well as additional features to support collaboration and the secure sharing of research results.

The overarching objective remains to establish DeepLaw in the long term as the central, reliable research platform for Swiss law and to sustainably enhance the efficiency of legal work.