The growing use of artificial intelligence (‘AI’) coupled with the rise of open data of courts’ decisions provide us with the tools to analyse and predict judicial outcomes [1].
Legal techs — startups developing a range of dematerialised legal services — are taking over the market of legal prediction. Legal prediction is the analysis of a large dataset of court decisions using artificial intelligence technologies to forecast the outcome of disputes. To be efficient, legal prediction tools are fed with decisions rendered by courts from the lowest to the highest courts.
France has enshrined the opening of public data with the Digital Republic Act 2016 [2]. This Act makes it easier to have access to courts’ decisions online. If one decision is not published, third parties can request to have access to the court’s decision. It enables legal techs to have an extensive and growing database. However, this trend might be hampered by an Application Decree of June 2020 [3], which details the availability to the public of the judicial and administrative courts’ decisions. One provision in particular restricts the delivery of a copy of a court’s decisions to third parties [4]. This restriction raises concerns about the prosperity of legal techs.
Indeed, legal techs collect courts’ decisions and process the data to have a solid database, which then enables the machine to predict the outcome more accurately. This is only possible thanks to open data and unlimited delivery of courts’ decisions.
Legal prediction seems to push our society to have a transparent judiciary system (I) but the Decree implemented in June 2020 challenges this development (II).
I – The Act for a Digital Republic: transparency of justice thanks to open access and open data
The provisions of the Digital Republic Act provide that a copy of a court’s decision may be issued by the clerk’s office to any third party upon request. This is called ‘open access’, which is to be differentiated from ‘open data’. Open data is the dissemination of case law data in electronic form, free of charge, thanks to the website Légifrance. Predictive justice was encouraged by French policymakers with the Senate report « Five years to save justice » published in April 2017 [5]. The public has obtained a wider access to cases thanks to Légifrance over the years.
However, improvements still need to be made, as some cases, mainly lower courts’ decisions, are still not getting published and are therefore unavailable to the public.
As a consequence of this lack of publication, lawyers, legal departments, insurance companies and even litigants are turning to legal techs to optimize their legal strategies, by benefiting from the database of legal techs.
Indeed, thanks to the principle of open access of the Digital Republic Act, French legal techs were used to requesting the delivery of unpublished court decisions to the clerk’s office in order to collect as many court cases as possible, until the Decree of June 2020.
II – The scope of the Digital Republic Act restricted with the Decree of June 2020 relating to the availability to the public of the decisions of the judicial and administrative courts
The recent decree questions the future prosperity of French legal techs. Indeed, the perimeter of the Decree is rather limited in terms of collecting decisions. The problem is the following: abusive requests are not accepted, particularly because of their number or their repetitive or systematic nature [6]. Legal techs make repetitive requests to the clerks in order to collect all the courts’ decisions that have not been made public.
Even though the open data regime for courts’ decisions has become broader than before with the Act for a Digital Republic, the Decree is a limit to open access and a limit to the collection of data. Thus, legal techs will only be able to reuse the data within the framework of the Decree, which is not in their favour.
Pascaline BAILLIEUX
[1] Nikolaos Aletras, Dimitrios Tsarapatsanis, Daniel Preoţiuc-Pietro and Vasileios Lampos ‘Predicting judicial decisions of the European Court of Human Rights: a Natural Language Processing perspective’ (2016) 2(93) PeerJComput.Sci. 1
[2] LOI n° 2016-1321 du 7 octobre 2016 pour une République numérique
[3] Décret n° 2020-797 du 29 juin 2020 relatif à la mise à la disposition du public des décisions des juridictions judiciaires et administratives
[4] Ibid. Titre 1: Article 1 al 4. « Abusive requests are not accepted, particularly because of their number or their repetitive or systematic nature – Traduction : Il n’est pas fait droit aux demandes abusives, notamment par leur nombre ou leur caractère répétitif ou systématique. » (soulignement ajouté)
[5] P. Bas, E. Benbassa, J Bigot, FN Buffet, C Culkierman, J Mézard, F Zocchetto, ‘Cinq ans pour sauver la Justice/ Five years to save Justice’ Report of the Senate n°495 (2016-2017) <https://www.senat.fr/notice-rapport/2016/r16-495-notice.html>
[6] « Abusive requests are not accepted, particularly because of their number or their repetitive or systematic nature – Traduction : Il n’est pas fait droit aux demandes abusives, notamment par leur nombre ou leur caractère répétitif ou systématique.” (soulignement ajouté)