STF uses artificial intelligence to speed up actions – 03/12/2024 – Power

Like several courts across the country, the STF (Supreme Federal Court) is trying to develop AI (Artificial Intelligence) technology to speed up the analysis of cases and free up its collection.

The court will select an AI technology to create summaries of some of its resources, in an initiative that adds to other software already in use in other courts.

At the end of last year, Supremo received 24 prototypes from technology companies. The idea is to develop generative AI tools capable of summarizing two types of processes that come to court — extraordinary appeals and their grievances.

Generative AI is defined by the ability to create content, such as texts and images, based on “learning” with data provided to it.

“The STF provided participants with a set of data, with procedural documents — all public — necessary for the preparation of AI projects,” he told Sheetby email, the president of the court, minister Luís Roberto Barroso.

“I will analyze all these prototypes so that we can define the best path forward.”

The minister also cited interest in creating tools that bring together images, audio and videos, with legal data, to assist in the production of draft decisions, and the possibility of developing a single interface for the different systems adopted in the courts.

Even before the Supreme Court announced the new selection, the court had already implemented systems that it designates as AIs. Available since 2017, “Victor” is used in resource analysis to identify topics of general repercussion.

The court also uses the VictorIA systems (implemented last year) to identify cases on the same subject and group them automatically, and RAFA 2030, which supports the classification of actions according to the UN Sustainable Development Goals.

The trend towards robotization —with software classified as AIs, or other, simpler, automation ones— is not restricted to the Supreme Court.

Still at the highest levels of the Judiciary, the STJ (Superior Court of Justice) uses Athos, developed internally at the court in 2018. This system is used “in several internal projects to group processes, to search for another process by jurisprudence, by similarity” , explains Daniel Miranda, chief advisor of the STJ’s artificial intelligence department.

The STJ has already signed Athos transfer agreements with 39 courts — so that the courts can create their own technological solutions based on it.

In the instances below, other initiatives are added: a panel organized by the CNJ (National Council of Justice) in 2022 listed 53 courts with some AI initiative, at different stages of development; a new edition of this panel, being prepared by the Council, should indicate an increase in the use of these technologies, totaling 140 projects and 61 courts.

The body maintains Sinapses, a type of repository of models and data used in AIs by Brazilian courts. The platform controls the versions used in the courts and works as a means of contact between digital systems.

Not everything, however, falls under the concept of AI — some software in use in courts performs simple automations. The only two AI projects credited to the TJ-SP (São Paulo Court of Justice) in the 2022 CNJ panel, for example, are not yet considered as such by the court itself.

“What we have is this part of robotization that is very present, of software that can help us improve the progress of the process”, says Paula Navarro, judge assistant to the court’s president.

The two court projects on the 2022 panel are under joint development with USP (University of São Paulo). One of them deals with reading and classifying petitions; the other, analyzing duplicity of payment guides.

Advances in court automation have a positive aspect given the increase in demand on the Judiciary, assesses Rivana Ricarte, president of Anadep (National Association of Public Defenders).

The defender, however, highlighted the need to “maintain the human perspective with artificial intelligence” — a position similar to that of the legal representatives heard by the Sheetwhich expose fears about the scope of these new technologies.

According to Rafael Leite Paulo, assistant judge at the CNJ presidency, AI initiatives in use or development in the Brazilian Judiciary will not suggest decisions to judges. But these systems are used to process information in bulk.

“I needed to evaluate a set of processes, I use artificial intelligence to identify them. ‘Look, among these 100, 200 processes that you have pending, these 50 here deal with the same topic; these other 20 here have exactly this document pending’ “, he exemplifies.

The development of AI systems “has a very large positive impact, including with regard to the agility of processes”, says Paulo Brincas, national coordinator of Innovation and Technology at the Federal Council of the OAB.

But the lawyer cites a concern about the lack of transparency in the construction of these systems, and the little participation of lawyers in this process.

“I believe that, very soon, these artificial intelligence systems will replace the administration systems we have today”, continues Brincas.

His fear is that what happened with administrative software in the Judiciary will happen to AI: a profusion of programs without centralization, a kind of “tower of Babel”.

Solano de Camargo, president of the OAB-SP Privacy, Data Protection and Artificial Intelligence Commission, highlights that many solutions similar to what is seen in the courts are being implemented by the legal profession itself, in a scenario in which these technologies are becoming more accessible.

So far, according to the lawyer, there has been no intention to replace the judge in decision-making — but if that were to happen, it would be an ethical violation. “We have the right to be judged by human beings and convince the judge that he may have a superior opinion.”