The practice of law has always been exercised through language, spoken or written. By means of texts that narrate facts, advance arguments, and give life to positive law, attorneys fulfill their role as an essential function of justice—setting the jurisdiction in motion, otherwise inert, to vindicate their clients’ rights. Thus, the text is the space where facts are organized, rights are claimed, risks are quantified, and decisions are influenced.
Accordingly, the broader a lawyer’s vocabulary, the greater their capacity to express, in clear prose, the needs of those who seek their technical expertise. One may therefore say that the quality of legal writing reflects the quality of legal thought: the more sophisticated the available linguistic resources, the greater the attorney’s ability to convert complexity into clarity, without sacrificing the technical rigor the profession demands.
With the advent of generative artificial intelligence tools—such as ChatGPT (OpenAI) and Copilot (Microsoft)—capable of drafting, summarizing documents, and suggesting arguments in seconds, text remains the primary input of legal work, but it is now partially produced by a non‑human agent. In this setting, the decisive point for the ethical practice of law is not merely the efficiency achieved, but how that efficiency is obtained and which professional duties remain entirely non‑negotiable.
These duties are not new to the legal profession, which has long been guided by the Brazilian Bar Statute (Law No. 8,906/1994) and the Code of Ethics and Discipline of the Brazilian Bar Association (CFOAB Resolution No. 02/2015). Together, they establish a normative core of professional conduct grounded in preserving the honor and dignity of the profession; acting with independence, honesty, decorum, truthfulness, loyalty, and good faith; and maintaining a permanent commitment to technical improvement and high‑quality delivery of legal services.
In light of this new professional context, the Federal Council of the Brazilian Bar Association issued specific guidance (Recommendation 001/2024) on the use of Generative Artificial Intelligence in legal practice. It acknowledges that available tools bring significant benefits, but also inherent risks that cannot be overlooked, and starts from the premise that there are acts reserved to attorneys which cannot be delegated to third parties—whether to unlicensed individuals or, in this new reality, to a non‑human agent.
Thus, Generative AI must clearly be understood as an auxiliary instrument under the lawyer’s technical judgment, not as an autonomous instance of legal decision‑making, precisely because responsibility for the content produced, selected, and presented remains fully with the attorney who signs it. It is equally recognized that Generative AI may produce erroneous, imprecise, or biased outputs, including due to poor or skewed training data, with discriminatory potential and concrete risks to clients and third parties, as well as reputational risk to the professional.
At this point, it is worth noting that an AI system functions essentially as a statistical model trained on large volumes of data, capable of recognizing patterns and, from them, generating a response when prompted. In the case of large language model (LLM)–based Generative AI tools, the core mechanism is as follows: the user’s text is “tokenized” into smaller units, and the system computes—based on the context supplied—which tokens are most likely to appear next, producing text sequentially until the response is complete.
Consequently, for AI use to yield the expected efficiency, the attorney’s writing ability becomes even more pivotal, as it is what converts the need for support into clear, complete, and technically bounded instructions—providing the tool with the context, objectives, and constraints necessary to generate outputs better aligned with the concrete case. In practical terms, output quality tends to mirror input quality: vague instructions will yield generic, and at times inaccurate or even useless, texts.
As a result, a lawyer’s familiarity with different narrative styles, pacing, and syntactic constructions enhances their capacity to choose words precisely, organize ideas coherently, and construct persuasive, lucid reasoning. In a professional environment where information is abundant and technology can accelerate text production, excellence will continue to depend on the lawyer’s cultural formation and critical discernment—which is why reading habitually and writing methodically are not ancillary habits, but structural requirements of a responsible and innovative legal practice, in which every line drafted simultaneously conveys technique, strategy, and ethical duty.
Moreover, the ethical use of Generative AI in lawyering requires adopting a compliance model that combines confidentiality, diligence, truthfulness, human supervision, and transparency, ensuring that technology is incorporated as a means of improving legal services, not as a surrogate for the professional.
From an individual perspective, as society changes its means of communication, standards of access to information, and expectations regarding efficiency and transparency, the practical demands of professional performance are also reshaped—without altering the core duties that underpin them. Hence, continuous upskilling and the enhancement of writing, argumentation, and work methods become conditions for keeping pace with the present.
For these reasons, continuous study must be understood as a permanent and inescapable commitment, as it is what enables the attorney to accompany the changes of their time without relinquishing rigor, ethics, or professional excellence.
Available at: https://analise.com/opiniao/a-escrita-juridica-na-era-da-inteligencia-artificial?sep=analise
Autor: Raissa Varrasquim Pavon • email: raissa.pavon@ernestoborges.com.br
