Human control is still part of the responsibility in using AI
A lawyer can use artificial intelligence to summarize documents, research arguments or prepare a first version of an analysis. The technology can speed up these steps. But whoever answers for the content presented to the client, the interpretation of the case and the legal decision is still the professional.
This distinction was at the center of recent debates promoted by the Federal Council of the OAB, the Brazilian Bar Association. In August 2026, during a series of sessions dedicated to the topic, the organization highlighted integrity, human control, information quality, transparency and data protection.
AI can support reasoning without taking on responsibility
Generative systems can produce convincing texts, organize large volumes of information and suggest paths quickly. That doesn't mean they understand the case the way a professional does.
It's up to the lawyer to analyze evidence and documents, choose the grounds, build the argument and check whether the content produced matches reality. The greater the consequence of an application, the less appropriate it is to treat AI's first answer as the final version.
Human control needs to happen during the process
The recommendation isn't just reviewing the document before sending it. In the OAB debates, the case was made for substantial and permanent human control, following use from the choice of tool to the evaluation of results.
In practice, this requires checking whether the facts cited are real, whether case law and references exist, whether documents were interpreted correctly, whether important information was left out and whether the conclusion makes sense given the concrete case.
This care matters because AI models can produce incorrect information with an appearance of certainty.
Confidentiality and data protection come before the tool
Recommendation No. 1/2024 of the OAB Federal Council guides the use of generative AI considering legislation, confidentiality, privacy, professional responsibility, verification of information and communication with clients.
For a law firm, this requires understanding what happens to the data entered into each solution. Client documents, litigation strategies, personal data and other protected information shouldn't be sent to just any tool simply because it makes a task easier.
Gaining speed doesn't change who answers for the work
Artificial intelligence can reduce part of the manual execution and expand the capacity of professionals and teams. But the gain doesn't transfer responsibility to the system.
Before expanding use, the firm needs to define in which activities the technology can help, what information can be entered, which results require mandatory review, who takes the final decision, how errors will be identified and when the client needs to be informed.
The goal isn't to limit the use of AI. It's to make clear how far the tool goes and where professional responsibility begins.
Sources
Conselho Federal da OAB, 25 Aug 2026. https://www.oab.org.br/noticia/64511/forum-nacional-de-inteligencia-artificial-na-advocacia-encerra-programacao-com-compromissos-de-formacao-protecao-e-inclusao
Conselho Federal da OAB, 27 Aug 2026. https://www.oab.org.br/noticia/64520/controle-humano-e-protecao-de-dados-pautam-3-dia-da-jornada-iv-da-oab-sobre-ia
Conselho Federal da OAB, 14 Nov 2024. https://diario.oab.org.br/pages/materia/842347
Plataforma de Eventos da OAB Nacional, 24 Aug 2026. https://centraleventos.oab.org.br/event/1546/advocacia-4-0-forum-nacional-de-inteligencia-artificial-na-advocacia-oab--stanford--idp
Plataforma de Eventos da OAB Nacional, 24 Aug 2026. https://centraleventos.oab.org.br/event/1514/jornada-iv-advocacia-inteligencia-artificial-e-novas-tecnologias