How AI Is Changing the Law

Portuguese version of this article here

Law firms, companies and courts in perspective

Brazilian law is changing. Not as a promise, but as a verifiable fact. The 2024 AI in the Judiciary Survey published by the CNJ recorded 178 artificial intelligence projects, 98 of them new initiatives. Those figures came from 63% of Brazilian courts, with 96.8% of judicial bodies taking part.

In April 2026, the CNJ went further still. At IAJus 2026, it launched Sinapses 2.0 and unveiled the Berna tool, which analysed 30 million cases across 88 courts to identify patterns of abusive litigation. AI has moved well beyond the experimental margins.

This piece builds on our recent article on the AI battle in Brazilian courts, taking the analysis a step further, how is this transformation playing out in practice, inside law firms and legal departments and where do the risks still outweigh the gains?

Artificial intelligence in the Brazilian Judiciary

Source: CNJ - AI survey / berna tool, March 2026

Judiciary in Transformation

Brazil entered the digital era at the top of international rankings for open data, internet usage and data centre investment. That digital foundation has had direct consequences for the legal environment.

Paper files and manual workflows have given way to digital case management, virtual hearings and automation. The Supreme Court has pressed ahead with the Maria platform and the Superior Court of Justice with Logos, both designed to triage caseloads that would be unmanageable without automation.

In March 2025, CNJ Resolution 615/2025 set out the main regulatory framework for AI in the judiciary. It requires human oversight at every stage, prohibits fully automated rulings and establishes the National Committee on Judicial AI.

The resolution draws on the model of the European AI Act, in force since August 2024, though the two contexts have taken different paths. In Europe, regulation preceded widespread adoption in the justice sector. In Brazil, the tools and the rules have been developing in parallel.

Professor Paula Costa e Silva, of the University of Lisbon, follows the Ibero-American judicial system closely. For her, Brazil's edge is not purely technological.

“The intelligence is grounded in solutions and data instruments, with a rational, well-structured approach to understanding problems. Brazil has already shown it can map challenges and develop highly appropriate responses by comparison with other countries.”

Paula Costa e Silva, Professor, Faculty of Law, University of Lisbon, and practising lawyer in Portugal and Brazil

What stands out, she argues, is Brazil's ability to generate data in real time. When a wave of litigation begins to build, it is possible to respond quickly and contain the problem before it compounds. In systems where data arrives late, the cost of reacting is always higher.

Inside the Law Firms

Guilherme Rizzo Amaral, founding partner at Souto Correa Advogados, draws a distinction between three separate actors: law firms, arbitrators and the courts. Within law firms, AI is already an established reality. Around 94.4% of professionals at his firm use the technology on a regular basis, with Copilot integrated across the firm's main software platforms.

“The impact on reviews, repetitive tasks and translations has been remarkable. We use it extensively for case law and legal doctrine research, with strong results. AI agents for specific proceedings, real-time cross-examination support, these are genuine advances. That said, handing over legal work entirely to AI is something I would strongly caution against.”

Guilherme Rizzo Amaral, Partner, Souto Correa Advogados

In arbitration, he notes, the technology tends to be used with greater caution and closer supervision. The picture in the courts is more troubling: decisions risk becoming excessively standardised, failing to engage with the specific facts of a case, and produced without adequate human review.

Tônia Russomano Machado, partner and head of the labour and employment practice at Andrade Maia, is careful to distinguish between automation and artificial intelligence. Long before AI arrived, the firm was already using automated tools for operational tasks, case tracking, system updates, report generation. AI, she argues, represents a step change. It extends the capacity for analysis and interpretation rather than simply executing pre-defined instructions.

In labour and employment law, the reach of jurimetrics has expanded considerably. Andrade Maia now produces structured dashboards covering average claim values, cost savings, market benchmarks and settlement timing, alongside contingency analysis, litigation risk mapping and early identification of recurring patterns. Much of that work used to be done manually, relying on intuition as much as data. Today, decisions are supported by analytics and predictive modelling.

“Above all, AI has raised the quality of what we deliver. The aim was never to replace people, but to free our professionals from ancillary tasks so they can focus on the work that matters most. What has grown is the standard of service we are able to offer clients.”

Tônia Russomano Machado, Partner, Andrade Maia

The firm's next move is the development of its own AI system, conceived as a firm-wide institutional project across all practice areas, bringing together the best of existing tools and adapting them to the specific needs of each discipline.

AI as Competitive Advantage

For companies with in-house legal teams, questions of governance take on an added dimension. The impact on the workforce is tangible and demands clear policy before any technology decision is made.

Alessandra Mizuta de Brito, Legal Coordinator and Data Protection Officer at Cattalini Terminais, is at the centre of that process.

“AI implementation should not be assessed purely through the lens of technology or efficiency. Responsible governance means considering the impact on people as well. Upskilling, reskilling and redeployment all need to be part of that conversation, a point already reflected in the Brazilian regulatory debate around Bill 2,338/2023 and in international references such as the European AI Act.”

Alessandra Mizuta de Brito, Doctor of Law, Legal Coordinator and DPO, Cattalini Terminais Marítimos S.A.

Larissa Alves Bastos Pedro, head of legal at TAP in Brazil, brings a different angle: using AI to manage mass litigation in the aviation sector.

“The scale of litigation in Brazilian aviation demands tools that can handle volume and pattern recognition simultaneously. AI helps identify repetition, structure strategy and anticipate trends, without replacing the legal judgement required for each individual case.”

Larissa Alves Bastos Pedro, Head of Legal, TAP Brazil

She also highlights the InfoVoo application, which consolidates data on delays and cancellations to support judges in aviation-related decisions. For airlines, the tool provides greater clarity on the real causes of disruption and what constitutes force majeure, giving the courts a more reliable evidential basis on which to rule.

These are not isolated examples. They point to a systemic capacity for data mapping, institutional co-operation and timely response that places Brazil in a strong position when it comes to the practical application of AI in law.

What Comes Next?

The current landscape is developing at three distinct speeds. The judiciary is operating AI at scale, deploying its own tools in pursuit of greater efficiency. Law firms that have approached the technology methodically are seeing real returns, not just in process automation but in the depth of analysis they can offer. In companies, governance is still catching up with the pace of change, though the benefits are already visible, particularly in litigation management.

What these three realities share is telling. The technology works well where someone is accountable for it. Where there is protocol, oversight and a culture of review. In short, where the final decision remains a human one.

The risks are not hypothetical. A recent study by FGV Justiça identified hallucinations in language model outputs, manipulation through prompts embedded in legal documents, algorithmic bias and the growing phenomenon of shadow AI, the use of private tools outside any governance framework.

The next step is not a technological one. It is cultural. Those who understand that will not simply use AI better. They will set the standard for what good use looks like.