Jus Mundi raises 8.5 million euros to establish its AI in international law
Case law, treaties, charters. The Paris legaltech offers free access to a database of 43,000 documents centred on international arbitration. A goldmine for lawyers.
Founded in 2019 in Paris, the legaltech Jus Mundi announces that it closed a second funding round at the start of September. Worth 8.5 million euros, the deal is led by C4 Ventures. Every backer who took part in the first round of 1 million euros last April is putting money in again, among them Holnest, INSEAD BA, Seed4Soft and Elkstone. They are joined for the occasion by Financière Saint James and the American fund FJ Labs.
Jus Mundi builds its offering around a base of 43,000 documents centred on international arbitration law. It covers case law, agreements and treaties, all freely accessible. On top of that comes a range of paid AI-based knowledge management services: access to more than 20 filters to combine search criteria (by type of case, sector, seat of arbitration and so on), and cross-links between content according to the legal references mentioned, including in other languages.
"Our natural language processing (NLP) environment lets a lawyer save up to five hours of work per arbitration case," says Thomas Latterner, CTO and co-founder of the startup. Behind the scenes, Jus Mundi's technical team relies on the open source software library PyTorch created by Facebook to build its artificial neural networks. Within the NLP libraries of the PyTorch galaxy, it draws in particular on spaCy, designed to process texts in 64 languages.
Mapping case law
Upstream, legal content is gathered from partner institutions, chief among them the International Chamber of Commerce (ICC) and the International Bar Association (IBA). Legal databases published as open data are also scraped. The NLP models then come into play to detect key legal concepts: "named entities" that go on to feed filtered search and weave the map of references between documents (what Jus Mundi calls its CiteMap). The whole structure rests on the open source indexing engine Elasticsearch.
"The challenge is to distinguish, among hundreds of paragraphs, those relating to the tribunal's reasoning"
"We are integrating automatic paragraph type detection into our deep learning models," says Benjamin Clavié, lead NLP at Jus Mundi. "The challenge is to distinguish, within arbitration decision documents, the paragraphs relating to the tribunal's reasoning from those linked to the history of the procedure, the parties' explanations and the rest. Given that these documents can run to 200 pages and between 600 and 1,000 paragraphs, the time saved will be considerable for lawyers."
Another R&D project under way, again around NLP: a neural network to anonymise the database while preserving the coherence of the text. "It identifies people's names, in whatever form, so as to replace them with a code: person 1, person 2 and so on," says Benjamin Clavié. For this project, Jus Mundi will follow its usual methodology: develop, test and then validate the model in English first, before adapting it to other languages.
Towards the creation of a legal graph
Alongside its data platform, Jus Mundi has built a directory listing 25,000 international arbitrators, lawyers and other experts in international law. Largely freely available, it again reserves certain key information for paying subscribers. The first of those? The complete list of cases handled by each professional, with all the associated decisions and documents. Better still, the premium version of the directory goes as far as identifying the practitioners and organisations they work for (or have worked for) who were involved in the same cases. A second search engine, called Conflict Checker, lets you reach all of an indexed lawyer's connections from their name. "We are migrating this application to the graph database Neo4J," says the startup's CTO.
Jus Mundi has just passed one million in annual recurring revenue (ARR), five times more than a year earlier. Unsurprisingly, 90% of the business is done abroad, 40% of it in the United States. "67% of the largest law firms are American, and 30% English," notes Jean-Rémi de Maistre, co-founder and CEO of Jus Mundi and a former international lawyer. Alongside North America and Europe, the company also sells its offering in Latin America and Asia.
Among its clients, Jus Mundi counts around a hundred law firms, including the prestigious Curtis, Dentons, DLA Piper, Eversheds Sutherland and Freshfields. But also central administrations such as the French ministry of foreign affairs and the British Foreign Office. Not forgetting universities such as Cambridge, Harvard and the Sorbonne.
A diversification strategy
The market Jus Mundi targets offers fine growth prospects. "Two thirds of American or European companies with annual revenue above 100 million euros have at least one international arbitration under way, and 40% have more than four," reckons Jean-Rémi de Maistre. The main advantage? Legaltechs are still very rare in this market.
To keep its lead, Jus Mundi intends to diversify its offering beyond arbitration and extend into sports law, environmental law and international taxation. On the commercial side, the company is preparing to open its first office abroad. "It will be based in New York to better target our flagship market," says Jean-Rémi de Maistre.
From 40 employees today, Jus Mundi intends to reach 80 people by the summer of 2022. Half the recruits will join the sales force, the other half R&D, targeting developer, data scientist and NLP expert profiles.