Artificial Intelligence and Digital Civil Law: Brazil's Civil Code Reform

Introduction

On January 31, 2025, Senator Rodrigo Pacheco introduced Bill No. 4/2025 in the Federal Senate, proposing a broad reform of the Brazilian Civil Code of 2002 (Law No. 10,406/2002). The draft originates from a preliminary text prepared by a commission of jurists convened during Pacheco’s presidency of the Senate and aims to update Brazil’s civil codification in light of the technological, social, and economic transformations of the past two decades.

The proposal, however, has already sparked intense debate. Critics argue that the draft was concluded with excessive haste and limited public participation, raising concerns about its theoretical depth and dogmatic consistency. Reforming what is often called “the second most important law of the country” — second only to the Constitution — demands careful deliberation, particularly when the text seeks to encompass phenomena as recent as artificial intelligence, digital identity, and neuro-rights.

In this context, and in light of the criticisms regarding the speed of the legislative process and the limited public participation in the drafting of the preliminary bill, the Federal Senate has sought to broaden its dialogue with the academic community and civil society. Within the scope of the Temporary Commission for the Update of the Civil Code (CTCIVIL), a series of thematic public hearings has been scheduled, among which stands out the debate on the General Part of the Civil Code and Digital Civil Law, set to take place on October 16, 2025.

Indeed, the incorporation of artificial intelligence regulation into Book VI – Digital Civil Law constitutes one of the most ambitious and, at the same time, most controversial movements of this reform. It represents an attempt to “civilize” the technological phenomenon — that is, to subject it to classical civil law principles such as good faith, human dignity, social function, and full compensation for damages. Article 2.027-AL of the bill introduces a civil liability regime for AI within the Code itself, establishing that every algorithmic operation must be legally anchored in a natural or legal person, in accordance with the principle of full reparation.

However, this proposal emerges at a moment when Congress is already discussing the Artificial Intelligence Legal Framework (Bill No. 2.338/2023), whose most recent version sets forth a technical and preventive regulatory regime, based on governance, risk management, and human oversight. The coexistence of these two initiatives — one civil and substantive in nature, the other regulatory and technical — raises a central question: what is the proper place of AI within Brazilian civil law?

While Bill No. 2.338/2023 defines how to prevent harm (through compliance mechanisms and risk management), Bill No. 4/2025 defines how to repair harm (through attribution of liability and the duty to compensate). The risk, however, lies in the absence of coordination between the two, which may lead to normative duplication and legal uncertainty, particularly if each statute is interpreted in isolation.

Comparative and Structural Approach (Bill No. 2.338/2023 vs. Bill No. 4/2025)

Bill No. 2.338/2023 represents the most comprehensive and advanced proposal for the regulation of artificial intelligence in Brazil. Drafted through an extensive process of legislative debates and public consultations led by the Senate’s Commission of Jurists between 2022 and 2023, it consolidated several previous proposals — notably Bills No. 5.051/2019 and No. 21/2020 — and was formally presented in May 2023 by Senate President Rodrigo Pacheco.

The bill’s central objective is to establish a national regulatory framework for the ethical and safe development, implementation, and use of AI systems. Inspired by the European Union’s AI Act, it adopts a risk-based approach, classifying systems according to their potential harm to fundamental rights and public safety. Applications are divided into three main categories: unacceptable risk, high risk, and low or minimal risk. Unacceptable-risk systems — such as those that use subliminal techniques to manipulate human behavior or engage in social scoring — are expressly prohibited.

High-risk systems, which include medical diagnostics, hiring and employment management, credit scoring, autonomous vehicles, public security, and criminal justice, are permitted but subject to strict requirements of transparency, documentation, and prior impact assessment. Low- or minimal-risk systems — such as spam filters, spell checkers, and content recommendation algorithms — are exempt from additional obligations.

From a comparative perspective, the Brazilian model is broader than its European counterpart: while the EU’s AI Act sets forth 39 obligations, the Brazilian proposal lists at least 69, raising concerns about compliance burdens for small enterprises and startups.

When contrasting Article 2.027-AL of Bill No. 4/2025 with Bill No. 2.338/2023, currently pending in the Chamber of Deputies, one observes structural differences in nature, purpose, and systemic integration.

Both projects share the same axiological foundation — that AI development and use must be subordinated to human dignity, non-discrimination, transparency, and accountability. Yet Bill No. 2.338/2023 establishes an autonomous regulatory framework, technical and sectoral in nature, while Bill No. 4/2025 seeks to integrate AI directly into the core of Civil Law, through the inclusion of a new book titled “Digital Civil Law.”

From Technical Regulation to Civil Discipline: Legislative Competence and Federalism in AI Governance

The crucial difference between the two proposals lies in their underlying legal logic. Bill No. 2.338/2023 establishes an administrative regulatory framework, oriented by public policy and by a regime of technical compliance obligations, structured around risk management and impact classification, delegating regulatory detail to sectoral authorities. By contrast, Bill No. 4/2025, by incorporating AI into the Civil Code, enacts a shift in legal nature: the subject ceases to be one of technical regulation and becomes part of civil law, governing private relations under principles of responsibility, consent, harm, and reparation.

This change has significant constitutional implications concerning legislative competence. Article 22, item I, of the Federal Constitution grants the Union exclusive authority to legislate civil law. Consequently, once AI is recognized as a civil law institution, States, the Federal District, and Municipalities may no longer enact laws on the subject. Subnational entities, however, retain administrative and technical regulatory competence, limited to implementing public policies, overseeing specific applications, and regulating operational aspects within their respective domains.

An illustrative example of the federative relevance of this issue is the case of the State of Goiás, which in May 2025 enacted the first state-level AI law in Brazil. The Goian legislation was widely praised for pioneering a regional framework for algorithmic governance based on three pillars: (i) the prioritization of open-source models, to reduce dependence on large platforms and enhance transparency; (ii) the inclusion of AI education in public schools, fostering critical and technical literacy; and (iii) the promotion of technological sustainability, by incentivizing the use of renewable energy in data centers.

These measures fall clearly within the administrative and policy-making competence of subnational entities, aimed at implementing public policies in education, innovation, and environmental sustainability. Constitutionally, such initiatives are legitimate under Articles 23 and 24 of the Constitution, which provide for shared and concurrent competences to protect the environment, promote education, and encourage regional technological development. The Goian law, therefore, does not legislate on “AI civil law,” but rather establishes public policy guidelines and administrative best practices.

Nevertheless, the approval of the Civil Code reform — as proposed in Bill No. 4/2025, which brings AI into the core of civil law — could substantially alter this framework. If civil law is redefined to include obligations, liability, and standards of lawfulness specific to AI systems, state laws that create substantive rules on rights, duties, or responsibilities related to AI use could be declared unconstitutional.

Thus, frameworks like that of Goiás — legitimate today as instruments of public policy — could have their validity compromised if they extend beyond technical regulation and into civil legal relations mediated by AI. Paradoxically, a reform intended to ensure unity and coherence in civil law could, in practice, undermine successful local innovations, constraining the ability of States to design and test context-sensitive regulatory policies.

Conclusion

The inclusion of artificial intelligence in Book VI – Digital Civil Law, as proposed by Bill No. 4/2025, represents one of the most ambitious and delicate undertakings in the history of Brazilian civil codification. By codifying principles and rules applicable to algorithmic agents, the project seeks not only to adapt the Civil Code to new technological realities but also to reaffirm the centrality of the human person in the digital environment, restoring to private law its ethical and protective functions amid the automation of social relations.

From a constitutional standpoint, transferring the subject to the civil law domain redefines the federal boundaries of AI regulation. By transforming artificial intelligence into a civil law category, the legislature attracts to the Union’s exclusive competence (Art. 22, I, of the Constitution) all matters concerning liability, legality, and the juridical effects of algorithmic acts. Consequently, state laws currently addressing ethical and transparent AI use — such as Goiás’s pioneering 2025 framework — will remain valid only to the extent that they operate within public policy and technical regulation, without intruding into private law.

Although this centralization may be necessary to ensure dogmatic unity and systemic coherence, it risks reducing the normative experimentation space of the States, which have acted as laboratories of regulatory innovation. The challenge for the federal legislator, therefore, lies in designing a cooperative model in which the Civil Code serves as the principled foundation of the legal system, while preserving subnational autonomy to promote policies of education, sustainability, and algorithmic transparency.

Ultimately, the consolidation of artificial intelligence as a subject of civil law should not signify the extinction of local initiatives, but rather their integration into a multi-level normative system, guided by the complementarity between administrative regulation and civil protection. In this sense, the new Civil Code may fulfill its historical vocation: to provide stability and coherence amid technological transformation, ensuring that, even in the age of algorithms, human dignity remains the foundation and the limit of the law.