# Cne Ia_Na_Educacao_Brasileira.Html
**Source:** https://en.ianaescola.com.br/publicacoes_relevantes/Atos_Normativos/CNE-IA_na_Educacao_Brasileira.html
**Language:** English
---
Ministry of Education
National Education Council
Plenary Council Support Division
National Education Council - SGAS, Av. L2 Sul, Quadra 607, Lote 50 - 70200-670 - Brasília - DF, - Asa Sul Neighborhood, Brasília/DF, Postal Code 70200-670
Phone: (61) 2022-7700 - http://www.mec.gov.br
| | | |
| --- | --- | --- |
| **INTERESTED PARTY:** National Education Council/Plenary Council | | **STATE:** DF |
| **SUBJECT:** Guidelines for the Use of Artificial Intelligence - AI in Brazilian Education. | | |
| **COMMITTEE:** Monica Sapucaia Machado (President), Celso Niskier and Israel Matos Batista (Rapporteurs), Cleunice Matos Rehem, Elizabeth Regina Nunes Guedes, Gastão Dias Vieira, Givânia Maria da Silva, Ludhmila Abrahão Hajjar, Mauro Luiz Rabelo and Paulo Fossatti (members). | | |
| **PROCESS NO.:** 23001.001062/2024-18 | | |
| **OPINION CNE/CP NO.: 15/2026** | **COLLEGIATE BODY: CP** | **APPROVED ON:** **1/9/2026** |
**I. REPORT**
**SECTION 1 - INTRODUCTION**
The incorporation of Artificial Intelligence - AI into contemporary educational systems is configured as one of the most relevant phenomena of technological transformations affecting education in the 21st century. The growing availability of these technologies expands possibilities for supporting teaching, learning, assessment, and educational management processes, but also introduces challenges related to the protection of rights, transparency, equity, intellectual autonomy, and the integrity of educational processes. In this context, its use is not an end in itself nor does it presuppose automatic benefits, and it must occur in a way compatible with the objectives of education, the protection of students, and the principles established in this Opinion.
These Guidelines are of a guiding and regulatory nature, establishing parameters for the incorporation of AI in educational systems and institutions to occur ethically, responsibly, proportionally, and pedagogically justified. Their purpose is not to authorize or indiscriminately encourage the use of these technologies, but to ensure that each application is based on a legitimate educational purpose, necessity, proportionality, human supervision, and protection of fundamental rights, preserving the centrality of pedagogical mediation and the formative purposes of education.
For the purposes of this Opinion, Artificial Intelligence is understood as the set of computational systems capable of processing data, making inferences, recognizing patterns, generating content, supporting decision-making processes, or performing other functions based on computational models. This concept covers, among other applications, predictive systems, generative systems, intelligent agents, digital tutors, and other technological architectures, recognizing that they present different purposes, degrees of autonomy, levels of transparency, auditability, and risk profiles. For the purposes of these Guidelines, the use of general-purpose AI systems is distinguished from that carried out by solutions designed or adapted specifically for educational purposes, and their use must always be analyzed in light of the pedagogical purpose, regulatory proportionality, human supervision, personal data protection, and potential impacts on educational processes and student rights.
In the Brazilian context, the expansion of the use of AI-based applications occurs in a scenario marked by inequalities in access to technologies and, above all, in the conditions for their qualified use. These differences reflect aspects related to available infrastructure, connectivity, the training of education professionals, the institutional capacity of educational networks, and the socioeconomic conditions of students. This reality highlights the need for normative frameworks capable of guiding the use of these technologies ethically, responsibly, and in line with the educational purposes established in the Federal Constitution of 1988 and the Law of Directives and Bases of National Education - LDB (BRAZIL, 1988; BRAZIL, 1996).
Overcoming these inequalities requires not only investments in technological infrastructure but also the strengthening of human and institutional capacities necessary for the critical, ethical, and pedagogically consistent use of AI. In this sense, the continuous training of education professionals, combined with the availability of adequate technical support, equipment, and connectivity, constitutes a strategic condition for technological innovation to contribute to expanding educational opportunities, avoiding the reproduction or deepening of existing inequalities.
The use of AI in educational processes has the potential to support the personalization of learning, expand access to educational content and resources, improve student performance monitoring, and foster more adaptive and inclusive pedagogical practices. These possibilities, however, must be understood in light of their limitations and risks, especially concerning personal data protection, system transparency and explainability, bias mitigation, and the preservation of student rights. It is also recognized that AI applications have distinct impacts on educational processes, ranging from predominantly instrumental uses to situations that influence evaluations, academic monitoring, institutional decision-making, and rights protection, a circumstance that justifies the adoption of a proportional regulatory approach based on the context of use (BRASIL, 2024; UNESCO, 2022; OECD, 2024).
In this scenario, teaching performance remains a structuring element of the educational process. It is the teacher's responsibility to ensure pedagogical intentionality, contextualization of content, careful selection of technological resources, and critical interpretation of information produced by Artificial Intelligence systems, ensuring that their use is subordinate to formative objectives and preserves the relational, ethical, and humanistic dimension of education (UNESCO, 2023).
The use of AI in educational processes must occur under permanent human supervision, interpretation, and responsibility. Even though these technologies can support planning activities, monitoring, personalization of learning, and organization of information, their use does not remove or replace the pedagogical responsibility of education professionals. Human mediation is an indispensable condition to ensure educational intentionality, contextualization, ethical judgment, and protection of students' rights.
Given this scenario, the National Education Council – CNE, in the exercise of its legal attributions, constituted a specific commission with the purpose of analyzing the impacts of AI in Brazilian education and proposing guiding directives for its utilization. The work developed sought to articulate legal foundations, technical references, and contributions from different actors in the educational field, including specialists, educational institutions, representative entities, and national and international organizations. This approach allowed for the construction of a well-founded normative proposal, capable of engaging with the complexity of the topic and the demands of the Brazilian educational system.
The opinion resulting from this process establishes guiding principles for the use of AI in educational systems and institutions, at all levels, stages, and modalities. Its objective is to offer parameters that ensure the incorporation of these technologies occurs in a manner consistent with educational principles, with the protection of student rights, and with the valorization of teaching performance, guaranteeing that technological innovation is subordinate to pedagogical purposes and does not override the educational nature of formative processes.
Furthermore, the opinion seeks to promote conditions for the use of AI to occur in a contextualized manner, respecting the specificities of the different levels of education, the characteristics of educational institutions, and the country's regional diversities. In this sense, it is recognized that the effectiveness of the guidelines depends on the ability of educational systems and institutions to interpret and implement them according to their realities, without losing sight of the principles that guide their application.
The purpose of this opinion is to contextualize and ground the regulatory proposal, contributing to its proper understanding and implementation by educational systems and institutions. Thus, it seeks to offer subsidies that favor the use of Artificial Intelligence in a responsible, transparent manner aligned with the quality of education, promoting the balance between technological innovation and commitment to the comprehensive development of students.
Underlying the guidelines presented here is the understanding that AI must be placed at the service of human development, learning, citizenship, and the integral education of students. Technological innovation only acquires educational meaning when it contributes to expanding human capacities, strengthening pedagogical bonds, and promoting opportunities for personal and social development.
**SECTION 2 – COMMISSION, HISTORY OF CONSTITUTION AND METHODOLOGY**
The preparation of the Guidelines for the use of AI in Brazilian education was conducted within the scope of the CNE, through a Bicameral Commission established by the Chambers of Basic Education – CEB and Higher Education – CES, with the purpose of analyzing the impacts of AI on educational processes and proposing guiding principles for its use in educational systems.
The establishment of this Commission is part of the exercise of the legal and regulatory duties of the CNE, especially regarding the formulation of national guidelines for education and the advising of the Ministry of Education (MEC) in the definition of public educational policies. In this context, the creation of a specific body to address AI reflects the institutional recognition of the strategic relevance of the theme and the need to address it in a structured, well-founded manner, coordinated with overall educational policies.
The Commission was composed of Councilors from CEB and CES, bringing together distinct institutional perspectives on the contemporary challenges of education at its different levels and modalities, namely: Councilors Monica Sapucaia Machado (CES/CNE), President of the Commission; Celso Niskier (CES/CNE) and Israel Matos Batista (CEB/CNE), Rapporteurs; Cleunice Matos Rehem (CEB/CNE); Elizabeth Regina Nunes Guedes (CES/CNE); Gastão Dias Vieira (CEB/CNE); Ludhmila Abrahão Hajjar (CES/CNE); Mauro Luiz Rabelo (CES/CNE); and Paulo Fossatti (CES/CNE), members, through CNE/CP Ordinance No. 30, of December 16, 2024, recomposed as follows: Councilors Monica Sapucaia Machado (CES/CNE), President, Celso Niskier (CES/CNE) and Israel Matos Batista (CEB/CNE), Rapporteurs, Cleunice Matos Rehem (CEB/CNE), Elizabeth Regina Nunes Guedes (CES/CNE), Gastão Dias Vieira (CEB/CNE), Givânia Maria da Silva (CEB/CNE), Ludhmila Abrahão Hajjar (CES/CNE), Mauro Luiz Rabelo (CES/CNE) and Paulo Fossatti (CES/CNE), members, through CNE/CP Ordinance No. 2, of January 31, 2025. The Commission also had the participation of MaximiliaDamas as an external researcher.
The composition of the Commission sought to ensure plurality and representativeness, integrating different experiences and views on Brazilian education, including academic, regulatory, and institutional perspectives. This diversity was fundamental for building a comprehensive approach to the use of AI in education, allowing for the consideration of the specificities of Basic Education and Higher Education, as well as the different realities of educational networks and institutions.
In the development of its work, the Commission held systematic meetings within the scope of the Plenary Council – CP, promoting the continuous analysis of topics related to AI and its interface with education. These meetings enabled the progressive construction of understanding on the subject, as well as the consolidation of the proposals that gave rise to the opinion.
As part of the qualified listening process, a technical meeting was held with experts from different areas of knowledge, including education, technology, and law, with the participation of representatives from teaching and research institutions, as well as governmental bodies. This moment contributed to deepening the discussions, especially regarding the ethical, legal, and pedagogical aspects related to the use of AI.
Additionally, a public hearing was promoted, with broad participation of representatives from the educational community, including educational institutions, sector entities, researchers, teachers, students, and members of civil society. This expanded participation space allowed the incorporation of different perspectives and contributions, strengthening the democratic and collaborative nature of the guideline drafting process.
In this process, the contributions of the participating institutions deserve to be highlighted, such as the MEC, the CNE, the University of São Paulo – USP, the Federal University of Rio Grande do Sul – UFRGS, the Federal University of Mato Grosso do Sul – UFMS, the Federal University of Rio de Janeiro – UFRJ, the Fluminense Federal University – UFF, the Farroupilha Federal Institute – IFFar, the Brazilian Institute of Capital Markets – Ibmec, Colégio Pedro II/EdutecIA Observatory, Alana Institute, Palavra Aberta Institute, PECK Institute for Digital Citizenship, Institute for Technology and Society – ITS, Brazilian Association of Isolated and Integrated Colleges – Abrafi, Brazilian Association of Higher Education Maintainers – ABMES, Union of Entities Maintaining Higher Education Establishments in the State of São Paulo – Semesp, STHEM Brasil Group/MetaRed, National Association of Graduate Students – ANPG, Harvard University, in addition to educational robotics companies, *streamers* and experts in education and technology.
The methodological process adopted by the Commission was characterized by a structured, participatory, and iterative approach, combining document reviews, institutional dialogue, public consultation, hearings, technical meetings, and the progressive consolidation of the contributions received. National regulatory frameworks related to education, personal data protection, the rights of children and adolescents in the digital environment, copyright, and fundamental rights were considered, as well as international benchmarks on ethics, governance, and the responsible use of AI, seeking to ensure coherence between educational policy and the current legal system.
At the same time, international references dealing with the ethical and responsible use of AI were analyzed, highlighting documents from UNESCO and the Organisation for Economic Co-operation and Development (OECD), which address topics such as transparency, fairness, governance, and the qualification of education professionals (UNESCO, 2022; OECD, 2024).
The systematization of the contributions received was carried out in a collegial manner, ensuring coherence among the legal foundations, technical references, and the demands of the Brazilian educational system. This process resulted in the consolidation of a normative proposal guided by criteria of responsibility, feasibility, and adherence to educational practices, avoiding both excessively prescriptive approaches and the absence of clear parameters.
Thus, the present opinion reflects a process of collective construction, based on evidence, legal frameworks, and dialogue with the educational community. Its elaboration sought to ensure technical consistency and institutional legitimacy, contributing to the formulation of guidelines that guide the use of AI in education ethically, transparently, and aligned with pedagogical purposes.
The dynamic nature of AI technologies requires that the guidelines established here be understood as references subject to continuous improvement. The rapid evolution of technological systems, regulatory frameworks, and scientific evidence related to the topic reinforces the need for permanent monitoring and periodic updating of educational guidelines associated with its use.
**SECTION 3 - PRINCIPLES**
The use of AI in Brazilian education will be guided by a set of principles that establish the regulatory foundations for its adoption, use, and monitoring in education systems and educational institutions. These principles express the conditions under which AI can be incorporated into educational processes, ensuring alignment with the purposes of education, current legislation, and national and international benchmarks on the responsible use of digital technologies (BRASIL, 1996; BRASIL, 2018; BRASIL, 2024; UNESCO, 2022; OECD, 2024).
The principles presented should not be understood in isolation, but as an integrated set that guides decision-making and the organization of AI-mediated educational practices. Their observance seeks to ensure that technological innovation is subordinated to pedagogical objectives, preserving the centrality of the educational process and the protection of the rights of the individuals involved.
The guiding principles are presented below, accompanied by a brief explanation:
**I - Centrality of human guidance with active pedagogical mediation**
It establishes that the use of AI must remain subordinate to the action of education professionals, ensuring that pedagogical decisions are guided by human judgment, educational intentionality, and institutional responsibility. This principle reaffirms the relational nature of education and recognizes that human mediation is indispensable for knowledge construction and the integral formation of students, especially in contexts mediated by technologies (BRASIL, 1996; UNESCO, 2023).
Human oversight encompasses responsibility for the conception, direction, monitoring, and evaluation of educational processes, including the selection, utilization, and critical review of AI-based tools. Its observance presupposes the preservation of human capacity to decide, interpret, contest, and intervene in the produced results, prohibiting the substitution of teaching performance or institutional responsibility by automated systems. Human mediation is not limited to the subsequent review of results produced by AI systems, but also includes the planning of activities, the definition of learning objectives, the selection of technologies, the monitoring of educational processes, and the critical evaluation of achieved results. The use of these tools must remain permanently subordinate to pedagogical intentionality and the responsibility of education professionals.
The centrality of human guidance and pedagogical mediation does not imply the individualized transfer to education professionals of technical, legal, administrative, contractual, or security responsibilities that belong to the education systems, educational institutions, maintainers, managers, technical teams, and suppliers of technological solutions. Teacher performance must be ensured under adequate institutional conditions, with training, technical support, work time, clear guidelines, pedagogical resources, and governance mechanisms that allow for the qualified exercise of human mediation.
The responsibility for the adoption, contracting, validation, monitoring, security, data protection, risk assessment, and institutional review of AI applications shall be clearly distributed among the different agents involved, preserving the pedagogical autonomy of education professionals and avoiding their accountability for systemic, technical, or contractual failures that are outside their sphere of control.
**II – Curricular integration of AI teaching, with progressivity and pedagogical intentionality**
It guides that teaching related to AI be incorporated into curricula in a planned, progressive, and pedagogically intentional manner, respecting the developmental stages of students and promoting meaningful learning. This integration should occur in conjunction with national digital education guidelines, current curricular bases, and the formative objectives of each stage and modality of teaching (BRASIL, 2017; BRASIL, 2025).
The curricular integration of AI should favor the progressive development of knowledge, skills, attitudes, and values compatible with the different stages of Basic Education and Higher Education, considering the specificities of the areas of knowledge and the developmental levels of students. This approach is not limited to mastering technological tools or applications, but also includes training for digital literacy, critical understanding of the foundations, potentialities, limitations, and impacts of AI in contemporary society, as well as the development of competencies for its ethical, responsible use aligned with educational objectives. Pedagogical intentionality constitutes a structuring element of this process, ensuring that technology remains subordinate to the integral development of students and the promotion of meaningful learning.
**III – Data protection, digital ethics, and information security**
It establishes that the use of AI must strictly observe the legislation related to the protection of personal data, as well as the principles of digital ethics, information security, and the protection of fundamental rights in the educational environment. This principle aims to ensure that the processing of information occurs in a lawful, transparent, secure, and responsible manner, preserving the privacy, integrity, and trust necessary for educational processes (BRASIL, 2018; UNESCO, 2022).
Its implementation requires the adoption of institutional policies and practices that guarantee adequate data governance, transparency regarding its use, the prevention of unauthorized access, and the mitigation of risks associated with the use of AI-based systems. Whenever applicable, the principles of data minimization, purpose limitation, retention for the necessary period, secure deletion, and anonymization or pseudonymization when appropriate must be observed, in compliance with current legislation. Educational institutions must seek to ensure that suppliers and other agents responsible for the development, provision, or operation of these solutions observe compatible standards of information security, personal data protection, transparency, and accountability regarding the processing of information used in educational contexts. The protection of personal data constitutes an inseparable dimension of the protection of the rights of students, teachers, and other members of the educational community, while digital ethics guides the responsible use of technologies, the prevention of discrimination, biases, and harms resulting from the automated processing of information, as well as the promotion of an institutional culture committed to privacy, security, integrity, and responsibility in the digital environment.
In the case of children and adolescents, the use of AI in educational contexts must also observe the legal regime of integral protection applicable to the digital environment, including specific standards aimed at the protection of children and adolescents in digital environments, especially regarding risk prevention, transparency, the protection of personal data, the prohibition of undue commercial exploitation, safety, human supervision, and the best interests of the child and adolescent, in accordance with the Digital Statute of the Child and Adolescent – Digital ECA.
**IV – Inclusion, equity, accessibility, and digital citizenship**
It establishes that the use of AI must contribute to expanding educational opportunities, reducing inequalities, and promoting equitable conditions for access, participation, and learning, strengthening inclusion, accessibility, and the formation of citizens capable of acting critically, ethically, and responsibly in digital culture (BRASIL, 2014; OECD, 2024).
The realization of this principle presupposes the recognition of the different conditions of infrastructure, connectivity, technological resources, and institutional capacities existing in the country. In this sense, education systems and educational institutions must adopt strategies that expand qualified access to technologies, promote the participation of all students, and reduce barriers associated with social, economic, territorial, cultural, or technological vulnerabilities.
AI may constitute an important instrument for promoting accessibility, the personalization of learning, the diversification of formative pathways, and the strengthening of conditions for permanence and academic success. Its use must be guided by the protection of rights, the valorization of diversity, and the commitment to ensuring that technological innovation contributes to expanding educational opportunities, rather than reproducing or deepening existing inequalities.
**V - Valuation of epistemic, cultural, and social diversity**
It establishes that the use of AI in education must recognize, respect, and value the plurality of knowledge, cultures, identities, contexts, and forms of knowledge production, promoting inclusive, democratic educational environments committed to diversity. This principle guides the critical use of technologies and the prevention of the reproduction of biases, stereotypes, or discrimination incompatible with human rights and the purposes of education (UNESCO, 2022).
The incorporation of AI in educational processes must consider the diversity that characterizes Brazilian society, recognizing different epistemological, cultural, territorial, and social perspectives as constitutive elements of knowledge production. In this context, AI systems and applications must contribute to expanding opportunities for participation, dialogue, and collective knowledge construction, avoiding the reproduction of homogeneous views, stereotypes, or patterns that might restrict the plurality of experiences, knowledge, and forms of learning. The appreciation of epistemic diversity is thus an essential condition for technological innovation to strengthen critical, inclusive, democratic, and socially committed education.
**VI – Pedagogical autonomy of educational institutions**
It ensures that educational institutions have the possibility to define, within the scope of their pedagogical projects and institutional specificities, strategies, policies, and forms of using AI compatible with their educational objectives and the needs of their academic communities. The exercise of this autonomy must observe the principles and guidelines established in this Opinion, as well as applicable legal frameworks, reconciling pedagogical innovation, institutional responsibility, and rights protection (BRASIL, 1996).
Institutional autonomy presupposes the capacity to develop contextualized solutions, suited to different educational realities, levels of education, and student profiles, favoring innovation processes aligned with the quality of education and the promotion of equity. Its exercise must occur in a transparent, participatory, and responsible manner, ensuring that decisions related to the use of AI remain subordinate to educational purposes, the integrity of formative processes, and the appreciation of the work of education professionals.
**VII – Scientific sovereignty and development of national solutions**
It encourages the strengthening of the country's scientific, technological, and institutional capacity for the development, evaluation, adaptation, and governance of AI solutions aimed at education, promoting the production of knowledge and technologies aligned with the needs, diversity, and strategic objectives of Brazilian education (BRASIL, 2021).
Scientific and technological sovereignty presupposes the strengthening of national ecosystems for research, innovation, and cooperation among educational institutions, research centers, the public sector, and the productive sector, favoring the creation and improvement of solutions compatible with the country's educational contexts. This principle stimulates the development of national capacities to understand, adapt, evaluate, and critically use AI technologies, contributing to reducing technological dependencies, expanding institutional autonomy, and ensuring that their incorporation into educational processes aligns with public interests, cultural diversity, and the purposes of education. The development and adoption of interoperable, auditable solutions compatible with open standards constitute a relevant guideline for strengthening scientific and technological sovereignty, while preserving the autonomy of educational systems and institutions.
**VIII – Algorithmic transparency**
Establishes that the use of AI in educational processes must be guided by appropriate standards of transparency, explainability, traceability, and documentation, compatible with the purpose, context of use, and risk level of the applications employed. This principle seeks to ensure that students, teachers, administrators, and institutions can understand the objectives, limits, and potential impacts of the technologies used, strengthening trust, accountability, and the protection of rights.
Algorithmic transparency comprises making accessible information available about the purpose of the system, its general operating logic, the data used, known limitations, and the criteria guiding its use in educational processes. Whenever applications influence assessments, academic monitoring, personalization of learning, or other decisions with a significant impact on students or institutions, proportional mechanisms for traceability, documentation, human supervision, and clear procedures for review, contestation, or correction of the produced results must be ensured.
**IX – Participation and social control**
It establishes that the use of AI in education must be guided by processes of participation, dialogue, and social control, ensuring the involvement of the educational community in the definition, implementation, monitoring, and improvement of policies and practices related to the use of these technologies. This principle reinforces the democratic management of education and institutional co-responsibility in building ethical, transparent digital environments committed to the public interest (BRAZIL, 1988).
AI governance in educational contexts shall favor the participation of students, teachers, administrators, researchers, families, and other segments of the educational community, respecting the specificities of each stage and modality of education. The incorporation of different perspectives strengthens the legitimacy of decisions, enhances institutional transparency, and contributes to the construction of contextualized solutions capable of responding to educational needs, keeping pace with technological transformations, and promoting the continuous improvement of teaching, learning, and management processes.
**X – Digital sustainability and environmental responsibility**
Establishes that the use of AI and other digital technologies in education must observe principles of sustainability, resource use efficiency, and socio-environmental responsibility, considering their impacts throughout the entire lifecycle of technological solutions. This principle recognizes that educational innovation must be aligned with sustainable development and the conscious use of natural and digital resources (UNESCO, 2023).
The adoption of digital technologies shall consider, in an integrated manner, their educational benefits, social and environmental costs, and the necessary conditions for their responsible and long-lasting use. Educational systems and institutions shall promote practices aimed at energy efficiency, the rational use of computing infrastructure, the proper management of the equipment life cycle, the reduction of electronic waste, and the responsible use of digital resources. At the same time, education shall contribute to the formation of citizens capable of understanding the relationships between technological innovation, sustainability, social development, and environmental responsibility, strengthening a culture of critical, conscious, and socially responsible use of technologies.
**XI – Digital well-being of students, teachers, and school managers**
It establishes that the use of AI and digital technologies in education must promote healthy, balanced learning environments that are conducive to the integral development of individuals, preserving the physical, mental, emotional, and social well-being of students, teachers, administrators, and other members of the educational community. This principle recognizes that technological innovation must serve human development and the quality of educational relationships, preventing adverse effects resulting from the inappropriate or excessive use of technologies (OECD, 2024).
The promotion of digital well-being presupposes the adoption of pedagogical and institutional practices that favor the conscious, critical, and balanced use of AI, preserving spaces for human interaction, intellectual autonomy, coexistence, creativity, and active participation in educational processes. Its observance involves attention to the cognitive, emotional, social, and behavioral impacts associated with the intensive use of technologies, as well as the prevention of informational overload, technological dependency, excessive surveillance, and other forms of use that may compromise health, learning, privacy, or the quality of educational relationships.
**XII – Intellectual autonomy and metacognition**
It establishes that the use of AI must contribute to the development of intellectual autonomy, critical reflection, and metacognition, helping students understand, monitor, and improve their own learning, reasoning, and decision-making processes. This principle reaffirms that digital technologies should act as instruments to expand human cognitive capacities, preserving individuals' creativity, critical thinking, authorship, and judgment capacity (UNESCO, 2023; OECD, 2024).
The integration of AI into educational processes must encourage practices that strengthen self-regulated learning, intellectual curiosity, problem-solving, collaborative knowledge production, and the conscious and responsible use of technologies. Pedagogical mediation should foster the development of competencies that enable students to critically analyze information, recognize limitations and biases of digital tools, formulate qualified questions, and use AI as support for research, creation, and knowledge construction, rather than as a substitute for intellectual effort.
In line with these objectives, the use of AI must strengthen intellectual authorship, academic integrity, and meaningful learning, and it is up to educational institutions to promote assessment practices that prioritize processes of investigation, critical reflection, problem-solving, projects, presentations, portfolios, and other strategies capable of demonstrating the effective development of students' competencies, autonomy, and intellectual responsibility.
**XIII - Federative cooperation, mutual technical support, and strengthening of the capacities of educational networks and systems**
Establishes that the implementation of guidelines related to the use of AI in education be guided by cooperation between federative entities, mutual technical support, and the strengthening of institutional capacities of educational systems and networks, respecting their different realities, competences, and levels of development. This principle reaffirms the collaboration regime set forth in the 1988 Federal Constitution and recognizes that the digital transformation of education constitutes a shared responsibility among the Union, states, Federal District, and municipalities (BRASIL, 1988).
The effectiveness of these guidelines presupposes the promotion of coordinated actions for the training of education professionals, knowledge sharing, the production of technical references, the development of institutional capacities, and the dissemination of good practices, favoring collaborative and sustainable solutions. The implementation of AI-related policies must consider the specificities of different educational networks and institutions, observing criteria of proportionality, equity, and federative cooperation, through technical support, the production of guidance, the sharing of experiences, and the development of institutional capacities, in order to expand local competencies and reduce inequalities in the implementation of these Guidelines.
The effectiveness of these guidelines presupposes the promotion of coordinated actions for the training of education professionals, knowledge sharing, production of technical references, development of institutional capacities, and dissemination of best practices, favoring collaborative and sustainable solutions. The implementation of policies related to Artificial Intelligence must consider the specificities of different educational networks and institutions, observing criteria of proportionality, equity, and permanent technical support, in order to expand local capacities and prevent structural inequalities from compromising qualified access to the opportunities offered by digital technologies.
The principles established herein constitute the normative foundation of the guidelines of this Opinion and guide their implementation by educational systems and institutions, respecting their specificities and autonomies. Guided by a humanistic perspective of education, they affirm that AI should act as a tool for expanding human and institutional capabilities, remaining subordinate to pedagogical mediation, the protection of fundamental rights, and educational purposes. Its integrated application seeks to ensure ethical, responsible, transparent, inclusive, sustainable, and proportional use in different contexts, contributing to the quality, equity, integrity, and development of educational processes.
**SECTION 4 - ADOPTION AND USE OF AI IN EDUCATION**
The adoption and use of AI in teaching systems and educational institutions should be understood as processes involving interdependent pedagogical, institutional, and technological decisions. In this sense, the incorporation of these technologies requires planning, grounding, and continuous monitoring, so as to ensure that their use is aligned with educational purposes and established principles.
The implementation of AI in the educational context cannot be reduced to the introduction of digital tools, but should be treated as an integral part of the organization of pedagogical work and institutional management. This perspective implies considering, in an articulated manner, aspects related to the quality of learning, the protection of rights, teacher performance, and the responsibility of educational institutions.
For purposes of adoption, governance, monitoring, and institutional evaluation, AI applications must be understood according to their predominant purposes of use. In this sense, applications aimed at pedagogical support, institutional and administrative management, evaluation processes, and accessibility and inclusion actions can be distinguished, among others. Although these categories may overlap in certain situations, their differentiation contributes to the definition of safeguards, responsibilities, and oversight mechanisms compatible with the risks and objectives of each context of use.
When a given application presents characteristics compatible with more than one purpose or risk category provided for in these Guidelines, the applicable criteria shall be cumulatively considered, with the classification corresponding to the highest risk level prevailing for the definition of safeguards, supervision mechanisms, and institutional governance. The classification shall consider the predominant purpose of use, the potential impact on students' rights, the degree of application autonomy, the nature of the data processed, and the relevance of the educational decisions supported or influenced by the AI system.
**4.1 Criteria for AI adoption**
The adoption of AI applications should be preceded by a well-founded technical, pedagogical, and institutional decision, which considers not only technological feasibility but also its adequacy to educational objectives and student needs. This process should involve the participation of the educational community, contributing to the legitimacy and transparency of decisions.
The analysis for adopting AI-based solutions should consider the nature of the intended application. Tools for pedagogical support, administrative management, learning assessment, or accessibility promotion may require differentiated criteria for analysis, supervision, and monitoring, especially when they produce relevant effects on educational processes or student rights.
Among the criteria that should guide this decision are:
- the pedagogical consistency of the adopted solutions and the existence of evidence of their effectiveness in supporting teaching and learning processes;
- the promotion of well-being and the integral development of students, considering cognitive, social, and emotional dimensions;
- the preservation of the autonomy and integrity of teaching work, avoiding the substitution of their role;
- the guarantee of security, equity, transparency, and accountability in the use of technologies and data;
- the analysis of the conditions of governance, storage, processing, and protection of the educational data involved, complying with current legislation and the strategic interests of Brazilian education;
- the verification of institutional capacity for the pedagogical use of AI, including the qualification of the professionals involved in its application;
- the evaluation of the conditions of technological infrastructure, connectivity, technical support, and operational sustainability necessary for the implementation of the adopted solutions; and
It is an analysis of the potential impacts of technology adoption on educational inclusion, considering different student profiles, institutional contexts, and regional realities.
The consideration of these criteria contributes to avoiding the adoption of technologies uncritically or out of context, ensuring that their use is effectively integrated into educational processes.
**4.2 Use of AI as a didactic resource**
AI can be used as a support resource for pedagogical practices, contributing to the diversification of teaching strategies, monitoring learning, and organizing educational activities. However, its use should always be complementary to the teacher's role.
Among the educational possibilities associated with the use of AI, the following stand out: support for the personalization of learning experiences, the expansion of resources for students with different educational needs, the strengthening of pedagogical monitoring strategies, and the provision of diverse ways of interacting with academic content and activities. Such applications must be subordinated to pedagogical objectives and teacher mediation.
The effectiveness of this use depends on the presence of clear pedagogical intentionality, which guides the selection, use, and evaluation of AI-based resources. In this context, teacher mediation plays a central role, ensuring the contextualization of content, the critical interpretation of the information produced, and the adaptation of activities to the students' needs.
The implementation of these guidelines must observe the principles of institutional proportionality and contextualized governance, considering the size, administrative capacity, technological infrastructure, and operating conditions of educational systems and institutions. In this context, a distinction must be made between the individual use of AI by students and teachers in diverse environments with limited capacity for institutional control, and the use promoted by educational institutions, guided by pedagogical objectives, teacher mediation, data protection, governance mechanisms, and training for the responsible use of technology. Educational policies should focus primarily on qualifying these institutional conditions of use, ensuring that governance requirements are compatible with the reality of each educational system, without prejudice to the observance of the principles of rights protection, transparency, human oversight, and the quality of educational processes.
The individual use of AI by the teacher, even if carried out outside the institutional environment or on their own initiative, is subject to the same conditions that preserve the proportionality between the teacher's professional autonomy and the protection of students and the integrity of educational processes.
The pedagogical use of AI must also consider the ethical and responsible management of these tools at all stages of their use, including the formulation of prompts, questions, and instructions, interaction with the systems, critical analysis of the responses produced, and the use of generated content. This perspective contributes to the development of intellectual autonomy, academic integrity, and critical awareness in the use of digital technologies.
In Distance Education - DE, the use of AI systems may support tutoring, guidance, learning monitoring, student support, and personalization of educational pathways, provided it is subordinated to institutional pedagogical planning, teacher curation, and significant human supervision. The use of these technologies cannot substitute the academic responsibilities proper to the teacher, tutor, or educational institution, and must contribute to expanding pedagogical interaction, accessibility, monitoring, and the quality of educational processes.
The use of AI tools in academic and educational activities must preserve the principles of intellectual authorship, academic honesty, and responsibility for the content produced. Interaction with AI systems does not transfer the responsibility for the presented content to the technology; it remains the full responsibility of the student, teacher, or researcher to assume authorship, veracity, consistency, and appropriateness of the information used or produced.
The integration of AI into educational practices therefore requires preparation for education professionals so that they can use it in a qualified, critical manner aligned with educational objectives.
Educational systems and institutions must promote ongoing training and updating programs for education professionals on the pedagogical use of AI. These programs may cover, among other aspects, algorithmic literacy, the creation of prompts and instructions for AI systems, critical curation of generated content, technology-mediated assessment, personal data protection, academic integrity, and the identification of algorithmic biases.
**4.3 Governance, documentation, and monitoring**
The use of AI in the educational context requires the adoption of institutional governance mechanisms that ensure transparency, control, and continuous monitoring. These mechanisms must allow for the identification of the purposes of the applications used, the analysis of the risks involved, and the definition of criteria for their evaluation and review.
Institutional governance mechanisms must consider the different purposes of the applications used. Systems employed for pedagogical support, academic management, evaluation processes, or the promotion of accessibility may demand specific procedures for documentation, monitoring, human validation, and institutional control, observing the proportionality between the risks involved and the potential impacts on educational processes.
Institutional documentation plays a fundamental role in this process by recording decisions related to the adoption and use of AI, as well as the parameters guiding its utilization. This record contributes to institutional accountability and the traceability of adopted practices.
The institutional governance processes must include mechanisms for the periodic review of policies, procedures, and criteria related to the use of AI, considering technological changes, regulatory updates, new scientific evidence, and the results observed in its implementation. Continuous updating is an essential element to ensure the effectiveness and adequacy of the adopted practices.
Faculty participation must also be present in the monitoring and supervision processes, ensuring that the use of technology remains aligned with pedagogical purposes and that any necessary adjustments are made in a well-founded manner.
**4.4 Transparency, communication, and human review**
The use of AI in educational processes must be accompanied by mechanisms that ensure transparency and understanding by the educational community. In this sense, it is necessary that the use of these technologies is communicated clearly to students, teachers, and administrators, allowing knowledge about their application and objectives.
The governance mechanisms applicable to the use of AI must observe the principle of regulatory proportionality, so that the levels of human supervision, transparency, documentation, traceability, and monitoring are compatible with the purpose, the context of use, and the risks associated with the applications employed. This approach seeks to ensure adequate protection for the rights of students and the educational community, avoiding both the imposition of disproportionate requirements for low-impact applications and the adoption of insufficient controls in situations capable of producing significant effects on educational processes.
Educational institutions must adopt measures to ensure adequate levels of explainability and traceability of the applications used, making it possible to understand their purpose, document the elements underlying recommendations, classifications, predictions, or generated content, and enable their review whenever there is a potential impact on educational decisions. In all these cases, the results generated by AI systems must remain subject to critical analysis and human responsibility, ensuring their pedagogical suitability, reliability, and compliance with the principles of this Opinion, as well as preventing the reproduction of errors, biases, or inappropriate information.
Human supervision, in this context, is not limited to the subsequent review of the results produced by AI systems. It also comprises the definition of pedagogical objectives, the selection of the applications used, the monitoring of their implementation, the continuous tracking of their performance, and the critical evaluation of the results produced, observing proportionality in relation to the educational purpose, the context of use, and the risk level of the application. Whenever there is a potential impact on rights, evaluation processes, or relevant educational decisions, a human review compatible with the nature and complexity of the technology's use must be ensured.
When human review involves education professionals, educational systems and institutions must ensure adequate conditions for its exercise, including specific training, institutional time, technical support, consultation channels, and decision-making backing. The ultimate responsibility for the governance of the application shall remain shared among the institution, the educational system, the sponsor, managers, and suppliers, according to their respective competences, and is not to be confused with the teacher's ordinary pedagogical responsibility.
**4.5 Limits on the use of AI**
The use of AI in education must observe limits that preserve the integrity of pedagogical processes and the formative nature of education. Among these limits, the impossibility of replacing teaching activities with automated systems stands out.
The preservation of the teacher's centrality ensures that the educational process remains grounded in human mediation, pedagogical interaction, and the meaningful construction of knowledge. In this sense, AI must be understood as a support tool, and not as a substitute element for the educational relationship.
Under circumstances may the use of AI result in the full transfer of pedagogical responsibility to automated systems. Processes related to learning assessment, academic monitoring, educational guidance, and decision-making affecting students' trajectories must remain subject to human supervision and validation compatible with their educational relevance.
**SECTION 5 – USE OF AI IN BASIC EDUCATION**
The use of AI in Basic Education must be guided by the formative purposes of this stage of education, considering the integral development of students and their age, cognitive, social, and emotional specificities. In this context, the incorporation of AI must occur in a planned, progressive, and integrated manner into the curriculum, avoiding fragmented approaches or those disconnected from educational objectives.
In Basic Education, the use of AI must observe the principles of integral protection of children and adolescents, age appropriateness, progressivity of learning, and the integral development of students, respecting the different stages of cognitive, emotional, social, and ethical development while preserving the centrality of human interactions in educational processes.
In Early Childhood Education, the total restriction of AI use by students is recommended, considering that this is a critical period for the child's integral development and that learning at this stage occurs primarily through responsive human interactions, affective bonds, play, and concrete exploration of the environment. Evidence gathered by the Scientific Committee of the Science for Infancy Center (2025) associates excessive screen use in the early years with impairments in language and communication, attention, emotional regulation, impulse control, motor development, and social interaction. Along the same lines, the Brazilian Society of Pediatrics - SBP recommends avoiding screens up to two years of age and strongly restricting their use between two and five years. In this context, the use of AI is not pedagogically justified and may compete with experiences fundamental to child development.
In the early years of elementary school, the use of generative and conversational artificial intelligence applications is not recommended, not even with teacher-mediated use. Additionally, at this stage, the use of AI systems must occur exclusively in pedagogical activities planned and mediated by education professionals, and their autonomous use by students is not recommended. The incorporation of these technologies must occur gradually, with a progressive expansion of their use as skills in digital literacy, critical thinking, digital citizenship, and intellectual autonomy develop.
In the Final Years of Elementary School, direct, autonomous, individual, or unsupervised access by students to generative AI tools will not be permitted, and their use is authorized only with the mediation of education professionals.
Age appropriateness should not be understood exclusively as a limitation of access to specific technologies. Considering the wide availability of AI tools in different digital, family, and social environments, the protection of children and adolescents depends, above all, on the conditions in which these technologies are used, with pedagogical mediation, human supervision, the protection of personal data, and the development of skills for their critical, ethical, and responsible use assuming special relevance. In this context, the presence of AI in Basic Education is not restricted to the use of technological tools, but also comprises the development of learnings that allow students to understand, use, and critically evaluate these technologies, articulating their pedagogical use with training for the exercise of citizenship in digital culture.
Whenever the institutional use of AI applications involves Basic Education students, educational systems and institutions must ensure transparency regarding the pedagogical purposes of the technology used, promote clear communication to families or guardians, strictly observe the protection of personal data, and prohibit the use of these applications for targeted advertising, marketing profiling, or commercial exploitation of student data.
**5.1 Integration of teaching relating AI to the curriculum**
The integration of AI into the Basic Education curriculum must occur in conjunction with digital, media, and computational education, respecting current national guidelines and the National Common Curricular Base – BNCC (BRAZIL, 2017; BRAZIL, 2023; BRAZIL, 2025). This integration should be carried out transversally and progressively, considering the different stages of teaching and the developmental characteristics of students.
This integration should observe the national curricular frameworks related to digital and computational education, promoting the progressive development of knowledge, skills, and attitudes compatible with each stage of schooling. The approach to AI should contribute to strengthening critical thinking, digital citizenship, digital culture, and the understanding of the relationships between technology, society, and democracy.
The curricular approach should favor the contextualization of learning, promoting the articulation between different areas of knowledge and stimulating interdisciplinarity. In this sense, AI can be treated both as an object of learning and as a pedagogical resource, provided that its use is aligned with the formative objectives and educational practices.
Furthermore, curricular integration must consider the reality of students and institutions, seeking to ensure that the use of technology contributes to the expansion of educational opportunities and not to the deepening of inequalities.
**5.2 Guidelines for the use of AI in Basic Education**
The use of AI in Basic Education must observe guidelines that ensure the quality of learning, the protection of students, and the pedagogical coherence of educational practices. As systematized in the guiding document on AI in Basic Education from the Ministry of Education (BRAZIL, 2026), the following aspects are highlighted:
the promotion of the integral development of students, considering cognitive, social, and emotional dimensions;
the use of resources that favor educational inclusion, accessibility, and the personalization of learning experiences, respecting the characteristics and needs of students;
Stimulus to the use of technologies capable of expanding opportunities for participation, engagement, and pedagogical monitoring;
Preservation of intellectual autonomy and student well-being, avoiding excessive dependence on technologies;
Integration of AI literacy in an articulated manner with digital, media, and computational literacies;
Contextualization of learning, based on students' reality and contemporary challenges;
Commitment to reducing educational inequalities, promoting inclusion and equity;
Consideration of students' and schools' digital access conditions, seeking pedagogical alternatives that avoid exclusion due to limitations in infrastructure, connectivity, or availability of technological devices;
Promotion of strategies that favor the democratization of access to educational opportunities supported by AI;
Development of pedagogical practices aligned with the protection of rights and digital citizenship;
Observance of age appropriateness for AI-supported educational tools, content, and experiences;
Strengthening students' ability to critically analyze information, synthetic content, and recommendations produced by automated systems;
Promoting safe, respectful digital environments compatible with the rights of children and adolescents.
Strengthening the critical and ethical mediation capacity of education professionals;
Preventing inadequate, invasive, or disproportionate monitoring practices of students.
These guidelines aim to ensure that the use of AI contributes to the formation of critical, autonomous individuals capable of understanding the role of technology in society.
**5.3 Development of AI-related learning**
Basic Education must promote the development of learning that allows students to understand the fundamentals, uses, and impacts of AI. This process involves not only the acquisition of technical knowledge but also the development of cognitive, ethical, and social skills.
Among the main aspects to be considered, the following stand out:
basic understanding of how computer systems and AI applications work;
recognition of different types of technologies and their uses in everyday life;
development of logical, analytical, and critical thinking;
reflection on the ethical, social, and cultural implications of technology use;
understanding the role of data and its use in AI systems;
understanding the basic foundations of AI, including its modes of operation, possibilities, limitations, and impacts on contemporary society;
understanding AI as a scientific, technological, economic, social, and cultural phenomenon that influences the organization of contemporary society;
identification of limits, risks, and possible biases present in technologies;
understanding the risks associated with disinformation, digital manipulation, synthetic content, and automated influence strategies present in digital environments;
- development of the ability to verify sources, validate information, and exercise critical judgment regarding content produced or mediated by AI;
- development of the ability to evaluate information produced by automated systems;
- understanding of the difference between technological support and independent intellectual creation, developing attitudes consistent with academic honesty and responsibility for the use of information produced or shared in digital environments;
- encouragement of creativity, autonomy, and problem solving;
- understanding of the relationships between technology, society, and inequalities;
- critical analysis of the social, economic, cultural, political, and ethical impacts associated with the development and use of AI;
- promotion of responsible, ethical, and safe behaviors in the digital environment;
The development of metacognitive capabilities related to the use of Artificial Intelligence, encouraging students to critically reflect on their own processes of thinking, learning, decision-making, and knowledge production mediated by digital technologies.
The approach to these learnings must foster interdisciplinarity and the construction of meaningful knowledge, contributing to the comprehensive development of students.
In Early Childhood Education and the Early Years of Elementary School, contact with concepts related to AI must occur primarily through mediated, contextualized, and developmentally appropriate pedagogical experiences, favoring exploration, creativity, problem-solving, and the progressive understanding of the role of technologies in daily life.
**SECTION 6 - USE OF AI IN HIGHER EDUCATION**
The use of AI in Higher Education must be guided by the articulation between academic training, scientific development, and preparation for professional contexts increasingly mediated by digital technologies. In this scenario, the incorporation of AI must be understood as an integral part of educational processes, contributing to the qualification of learning, pedagogical innovation, and the expansion of students' professional performance possibilities.
The diversity of knowledge areas present in Higher Education implies that the approach to AI will not occur homogeneously, and must consider the specificities of the courses, curricular guidelines, and institutional pedagogical projects. This diversity demands flexibility in the integration of AI, ensuring that its use is aligned with the educational objectives and characteristics of each area.
**6.1 Integration of AI into academic training**
AI can be integrated into academic training both as an object of study and as a support resource for teaching, research, and outreach activities. This integration must occur in a planned and articulated manner, promoting the connection between theoretical foundations, practical applications, and critical reflection.
The introduction of AI into academic training should contribute to the development of capabilities related to the analysis, interpretation, and use of data, as well as to the understanding of the impacts of technology in different professional fields. Furthermore, it should foster the training of professionals capable of acting in complex contexts, in which AI plays an increasing role in the production and organization of knowledge.
In undergraduate teaching degree courses, training related to AI should explicitly include teacher digital literacy, the pedagogical use of emerging technologies, the critical analysis of educational data, assessment assisted by digital technologies, and ethical aspects related to the use of AI in school environments.
Knowledge related to AI should, whenever possible, be articulated with supervised internship, extension, and research activities, favoring the approximation between teacher training and the concrete challenges faced by Basic Education institutions.
Higher Education Institutions - HEIs must establish their own policies and guidelines for the use of AI in teaching, research, extension, and academic management activities, encompassing principles of authorship, academic integrity, transparency, intellectual responsibility, and rights protection. These policies must, whenever relevant, be incorporated into the Institutional Development Plan - IDP, the Pedagogical Course Projects - PCP, bylaws, academic regulations, and other institutional governance instruments, ensuring coherence between institutional guidelines, pedagogical practices, evaluative processes, and academic management. They may include criteria for transparency regarding the use of technology, forms of declaration of use when relevant, guidelines for preventing practices incompatible with the principles of academic honesty, and procedures for monitoring, reviewing, and continuously improving institutional practices related to AI.
**6.2 Development of competencies related to AI**
Higher Education must promote the development of competencies that enable students to use AI critically, ethically, and responsibly, in line with the demands of contemporary professional contexts.
Among the main competencies to be developed, the following stand out:
• the ability to use AI as a support tool for decision-making in different professional contexts;
• the development of skills for data analysis, interpretation, and use;
• critical understanding of the possibilities, limits, and risks associated with the use of AI in learning assessment processes, academic tracking, and educational evidence analysis;
• understanding of the ethical, legal, and social implications related to the use of technology;
• critical evaluation of the limits, risks, and potentials of AI applications.
- comprehension of the conceptual foundations of AI and its impacts on knowledge production processes, work organization, exercise of citizenship, and social transformation;
- development of the capacity for critical and metacognitive reflection on knowledge production, learning, and decision-making processes mediated by AI;
- articulation between technical knowledge and reflection on the economic, social, and cultural impacts of technology;
- responsible use of AI, respecting principles of academic integrity and professional ethics;
- comprehension of the limits between technological support and one's own intellectual production;
- ability to use AI systems in a transparent and responsible manner in academic and professional contexts; and
- recognition of the importance of proper attribution of authorship and responsibility for content produced or presented with AI support.
These competencies contribute to the training of professionals capable of acting in a qualified manner in increasingly complex and technologically mediated environments.
**6.3 Pedagogical use of AI in Higher Education**
AI can be used as a support resource for pedagogical practices in Higher Education, contributing to the diversification of teaching strategies, the monitoring of learning, and the organization of academic activities.
The use of these technologies can contribute to the development of more flexible training pathways, the expansion of lifelong learning opportunities, the strengthening of student retention strategies, and the provision of academic support resources capable of meeting the diversity of profiles present in Higher Education.
Its use must be guided by a clear pedagogical intentionality, being integrated into the formative objectives of the courses and the teaching methodologies adopted. In this process, teacher mediation remains a central element, ensuring the quality of pedagogical interactions, the adequacy of the proposed activities, and the critical interpretation of the results produced by AI systems.
The pedagogical use of AI should avoid purely instrumental approaches, seeking to promote meaningful learning aligned with academic and professional demands.
**6.4 Ethical and Institutional Considerations**
The use of AI in Higher Education must observe ethical and legal principles, especially regarding the protection of personal data, transparency in the use of technologies, and respect for the rights of students and the academic community.
HEIs have a fundamental role in defining internal guidelines that guide the use of AI in its different dimensions, including teaching, research, and community outreach. This process must involve the academic community and consider the impacts of technology on institutional practices. The adoption of clear institutional policies contributes to the responsible use of AI, preventing risks and ensuring alignment with established principles and guidelines. Higher education institutions must establish their own policies and guidelines for the use of AI in teaching, research, outreach, and academic management activities, encompassing principles of authorship, academic integrity, transparency, intellectual responsibility, and the protection of rights. These guidelines may define criteria for declaring the use of technology when pertinent, guide practices compatible with academic honesty, and promote governance mechanisms aligned with educational and scientific purposes.
In the scope of scientific research, the use of Artificial Intelligence must observe the principles of scientific integrity, methodological transparency, and intellectual responsibility, and must be declared whenever its contribution is relevant to the drafting of texts, data analysis, bibliographic review, programming, or other stages of the investigative process. The use of these technologies does not transfer authorship or responsibility for the produced results to the Artificial Intelligence, which remain fully attributed to the researchers.
In academic management, AI may support planning activities, indicator analysis, identification of dropout risks, and the improvement of institutional processes, provided that significant human supervision is preserved, the possibility of decision review is ensured, and the adoption of exclusively automated decisions with a relevant impact on students' rights or academic trajectories is prohibited. In institutions offering teacher training courses, these policies must also include initial teacher training strategies, preparing future teachers to act critically, ethically, and pedagogically qualified in contexts increasingly mediated by digital technologies and intelligent systems.
**SECTION 7 – USE OF AI IN VOCATIONAL AND TECHNOLOGICAL EDUCATION**
The use of AI in Professional and Technological Education must observe the principles established in this Opinion, considering the specificities of this modality of teaching and its close articulation with scientific and technological development, innovation, entrepreneurship, and the world of work. Its incorporation must contribute to the comprehensive training of students, promoting the development of technical, digital, ethical, and socio-emotional skills necessary for critical, responsible, and innovative performance in professional contexts increasingly mediated by digital technologies.
The use of these technologies must favor active learning methodologies, project development, problem solving, experimentation, simulation of productive environments and the integration between theory and practice, expanding learning and professional development opportunities. Their adoption must be aligned with the formative objectives of the courses, the applicable National Curriculum Guidelines and the contemporary demands of scientific, technological and productive development, preserving the centrality of pedagogical mediation and the role of education professionals.
In courses involving laboratory activities, professional practices, internships, simulations, or the use of specialized equipment and environments, AI may constitute an important support tool for the training process, provided that its use does not compromise the development of practical competencies, professional autonomy, technical judgment, activity safety, and ethical responsibility inherent to the exercise of the professions.
The incorporation of AI in Professional and Technological Education should also contribute to strengthening the culture of innovation, applied research, creativity, lifelong learning, and the ability to adapt to technological transformations, preparing students to use these technologies in a critical, ethical, safe, and socially responsible manner, in line with the needs of sustainable development, national competitiveness, and the promotion of decent work.
**SECTION 8 – RISK CLASSIFICATION, SAFEGUARDS, AND REGULATORY PROPORTIONALITY**
The use of AI systems in the educational context does not occur homogeneously, since different applications present distinct levels of impact on pedagogical, administrative, and evaluative processes. In this sense, it becomes necessary to establish an approach guided by regulatory proportionality, considering that applications with greater potential for interference in rights, academic trajectories, or decision-making processes demand more robust mechanisms of supervision, governance, and institutional protection.
Considering the complexity and diversity of AI applications in the educational context, educational systems and institutions must adopt an approach guided by the principle of regulatory proportionality, according to which the levels of supervision, governance, documentation, transparency, and monitoring must be compatible with the educational purpose, the degree of autonomy of the system, and the risks effectively associated with its use, avoiding both the imposition of unnecessary requirements for low-impact applications and the insufficiency of safeguards in potentially more sensitive contexts.
Risk-tier classification helps differentiate between instrumental and pedagogical applications and systems with potential for significant impact on rights, academic trajectories, evaluation processes, and students' conditions for persistence. This approach is supported by international references on AI governance, especially in UNESCO recommendations and contemporary regulatory frameworks aimed at protecting fundamental rights and the responsible use of technology (UNESCO, 2022; UNESCO, 2023).
To facilitate the understanding and practical application of these guidelines, the risk classification of Artificial Intelligence applications must consider, in an integrated manner, its educational purpose, the degree of autonomy of the system, the type of data processed, the potential impact on students' rights, the level of decision automation, the possibility of profiling, and the reversibility of the effects produced. This evaluation must observe the principle of regulatory proportionality, according to which the level of supervision, governance, and institutional safeguards must be compatible with the risks effectively associated with each context of use.
Risk analysis should not be restricted to the technical characteristics of the solutions employed, but should also consider the concrete conditions of their use, including the existence of human supervision, pedagogical mediation, governance mechanisms, data protection, and institutional monitoring capacity. Thus, technologically similar applications may present distinct risk profiles when used for pedagogical support, institutional management, evaluation processes, promotion of accessibility, or other educational purposes, requiring contextualized analysis and the adoption of measures proportional to the potential impacts produced.
Restrictions related to the use of profiling mechanisms do not preclude the adoption of assistive technologies and resources exclusively aimed at promoting accessibility, educational inclusion, and the elimination of barriers to learning, provided that their use complies with the principles of data protection, human supervision, proportionality, and respect for students' rights.
In this regard, educational systems and institutions may consider, among others, the following risk levels for the purpose of organizing their governance, monitoring, and risk mitigation practices:
**I - Low-risk applications**
They correspond to applications for instrumental, organizational, complementary pedagogical, or accessibility support that, considering the purpose, context of use, nature of the data processed, and the system's degree of autonomy, do not produce a relevant impact on students' rights, evaluation processes, academic decisions, educational paths, or conditions of access, permanence, progression, and certification.
They can be understood as low-risk applications, among others, provided they do not involve individualized profiling, sensitive personal data processing, continuous monitoring, continuous interaction with students, or relevant automated decision-making:
a) organization, systematization, or adaptation of teaching materials;
b) auxiliary tools for pedagogical or administrative planning;
c) non-evaluative textual review;
d) support for translation, summarization, formatting, or content adaptation; and
e) simple accessibility features aimed at eliminating barriers, such as transcription, subtitling, screen reading, text enlargement, or format conversion, provided that they do not imply sensitive inferences, individualized profiling, or decisive repercussions on the student.
In these hypotheses, this Opinion guides that, as parameters of good institutional governance, duties of transparency, information security, personal data protection, institutional accountability, simplified record of the purpose of use, and the possibility of human supervision whenever the application is used in institutional educational activity should be observed.
**II – Moderate risk applications**
They correspond to applications used for pedagogical, academic, administrative, or institutional support that, although they do not produce relevant automated decisions, involve continuous interaction with students, generation of academic recommendations, personalization of learning paths, support for intellectual production, formative monitoring, institutional guidance, or systematic treatment of educational information.
Can be understood as moderate-risk applications, among others, provided that they do not produce automatic effects on grades, certification, progression, permanence, granting of benefits, disciplinary sanctions, selection, or other sensitive educational decisions:
a) automated tutoring or pedagogical support systems;
b) tools for *feedback* formative without direct academic repercussion;
c) institutional assistants used for service, general academic guidance, or informational support;
d) tools for supporting text production, research, or study organization, when used in a pedagogical or academic context;
e) academic recommendation mechanisms without automatic decision-making effect;
f) learning monitoring systems without direct and automatic repercussion on grades, certification, permanence, benefits, or sanctions; and
g) accessibility resources involving customization, individualized adaptation, or systematic processing of educational information, provided they do not involve sensitive data or relevant automated decisions.
In these cases, it is recommended that educational systems and institutions adopt safeguards proportional to the context of use, especially:
a) clear communication to the educational community about the use of technology, its purposes, limits, and institutional responsibilities;
b) institutional record of the applications used, indicating purpose, audience reached, data processed, and those responsible for monitoring;
c) mandatory human review of results whenever there is an influence on pedagogical guidance, academic monitoring, personalized learning paths, or recommendations directed at the student;
d) periodic monitoring of errors, biases, disproportionate impacts, and pedagogical adequacy;
e) prohibition of using educational data for training third-party models without an adequate legal basis, legitimate purpose, transparency, and compatible safeguards;
f) provision of an institutional channel for clarification, review, or contestation when the recommendation produced may affect the student's educational experience; and
g) periodic evaluation of the application's suitability to the originally assigned risk level, especially when there is an expansion of functionalities, integration with institutional databases, or a change in purpose.
**III – High-risk applications**
These correspond to applications capable of producing, supporting, or influencing significant effects on students' rights, academic life, evaluation processes, certification, selection, retention, progression, retention, granting of benefits, disciplinary analysis, distribution of educational opportunities, or other sensitive educational processes.
Must be understood as high-risk applications, among others:
a) automated support systems for the correction of assessments with academic repercussions, provided that they are used exclusively as an auxiliary instrument, with the automatic assignment of grades, concepts, approval, disapproval, certification, progression, retention, or any academic consequence without effective, prior, qualified, and documented human validation being prohibited;
b) systems of *proctoring* or other monitoring technologies used in assessment processes;
c) individualized academic profiling;
d) systems related to selection, certification, granting of benefits, permanence, progression, retention, or disciplinary analysis;
e) applications involving the processing of sensitive personal data, including data on disability, health, biometrics, behavior, social vulnerability, or sensitive individual performance;
f) applications involving relevant behavioral inferences, identification of individualized educational risk patterns, or prediction of dropout, performance, vulnerability, or conduct;
g) applications that may produce a relevant impact on educational opportunities, access to institutional resources, or conditions for continuing studies;
h) tools of *feedback* , tutoring, recommendation, or monitoring that begins to influence grades, progression, certification, permanence, benefits, sanctions, or other relevant institutional decisions; and
i) automated systems for pre-correcting essay evaluations, essays, and authorial textual productions, admitted exclusively as a preparatory and auxiliary instrument for the correction carried out by the teacher, who is solely and indelibly responsible for assigning the grade, concept, or merit and the formative feedback to the student.
In these scenarios, enhanced safeguards must be ensured, compatible with the sensitive nature of the application, including:
a) prior algorithmic impact assessment, proportional to the risk and educational purpose;
b) personal data protection impact report, when applicable;
c) significant, continuous, qualified, and documented human supervision;
d) mechanisms for traceability, documentation, and institutional auditing;
e) transparency and accessible explanation regarding the purpose, general operation, limits, and potential impacts of the application;
f) effective possibility of contestation, human review, and correction of results, recommendations, or decisions supported by AI systems;
g) prior and documented human validation of any result with evaluative, disciplinary, academic, administrative, or certificational repercussion;
h) periodic monitoring of biases, systematic errors, discriminatory or disproportionate impacts on specific groups;
i) adoption of less invasive alternatives whenever available, especially in evaluation, monitoring, individualized tracking, or sensitive data processing workflows; and
j) guarantee that the application does not replace the responsibility of an education professional or competent institutional body.
The use of high-risk applications may not result in an exclusively automated decision with a significant effect on the student's educational trajectory. The assignment of grades, certification, progression, retention, selection, granting of benefits, application of sanctions, and other relevant academic or institutional decisions must remain the responsibility of an education professional or competent institutional body, ensuring effective human supervision, justification, documentation, and the possibility of review.
**IV – Applications of excessive risk or incompatible with educational purposes**
These correspond to applications whose potential for rights violations, discrimination, undue surveillance, exploitation of vulnerabilities, or compromise of the integrity of the educational process makes them incompatible with educational principles and the integral protection of students.
They must be considered incompatible with these Guidelines, among others:
a) social scoring systems;
b) emotional surveillance or emotion recognition for evaluative, disciplinary, classificatory, or student conduct management purposes;
c) exploitation of student vulnerabilities, including age-related, emotional, socioeconomic, cognitive, or behavioral vulnerabilities;
d) psychological, biometric, or behavioral profiling for classificatory, disciplinary, commercial, or educational opportunity restriction purposes;
e) exclusively automated decisions with significant effects on promotion, retention, certification, retention in school, granting of benefits, dismissal, selection, or sanction;
f) continuous biometric surveillance or permanent monitoring of students;
g) use of educational data for targeted advertising, commercial exploitation, behavioral manipulation, or purposes incompatible with the educational relationship; and
h) systems that produce segregation, discrimination, or exclusion of students based on automated inferences about ability, behavior, vulnerability, or performance.
The potential use of point-in-time identity verification mechanisms, when strictly necessary for access to virtual environments, assessments, or academic activities, is not to be confused with continuous biometric surveillance, emotion recognition, behavioral profiling, or permanent monitoring, and must observe the principles of purpose, necessity, proportionality, personal data protection, transparency, human oversight, and the adoption of the least invasive alternative available.
Applications involving invasive biometric surveillance mechanisms, social scoring systems, or fully automated decisions with a relevant impact on educational rights or trajectories must be considered incompatible with the principles and educational purposes established in these Guidelines.
For illustrative purposes only, tools supporting the organization of teaching materials, non-evaluative textual review, and accessibility resources may be understood as low-risk applications; tutoring platforms, systems of *feedback* formative and academic assistants; and as high-risk applications, automated assessment systems with relevant academic impact, individualized educational profiling, and technologies used in sensitive decision-making processes. This classification is of an orientative and non-exhaustive character, and must be interpreted in light of the specific context of use and the principles of regulatory proportionality.
As a result of this classification, educational systems and institutions must adopt differentiated levels of governance, monitoring, and institutional safeguards. Low-risk applications may require simplified documentation and tracking procedures; moderate-risk applications must provide human oversight and periodic monitoring; and high-risk applications must be subject to enhanced mechanisms for impact assessment, traceability, human review, and rights protection.
The adoption of risk classification mechanisms does not alter the centrality of human action in educational processes, nor does it reduce institutional responsibility regarding decisions related to the use of AI. On the contrary, it reinforces the need for technological innovation to be subordinated to the protection of rights, pedagogical integrity, and the quality of education.
Furthermore, educational systems and educational institutions must maintain institutional governance mechanisms capable of ensuring continuous monitoring, periodic review, and updating of practices related to the use of AI, considering the rapid technological evolution and the associated educational, social, and ethical impacts.
Considering the continuous evolution of AI technologies, educational systems and educational institutions must periodically re-evaluate the classification of the applications used, as well as the adequacy of the safeguards and supervision mechanisms adopted, in light of the criteria established in these Guidelines, in order to ensure their adherence to technological transformations, new contexts of use, and the protection of students' rights.
**SECTION 9 - FINAL CONSIDERATIONS**
The incorporation of AI in Brazilian education is configured as a movement of a structuring nature, which demands the construction of a consistent balance between technological innovation, pedagogical responsibility, and the protection of rights. At the same time that it requires mechanisms for protection, supervision, and governance, AI offers relevant opportunities to expand inclusion, strengthen accessibility, diversify pedagogical strategies, and support more personalized learning processes. The responsible harnessing of these potentials constitutes one of the conditions for technological innovation to produce effective benefits for Brazilian education.
Throughout this opinion, it has become evident that the use of AI cannot be understood as an exclusively technological process, but as a transformation that directly impacts pedagogical practices, teaching and learning processes, assessment models, and forms of institutional organization. In this sense, the adoption of AI requires well-founded decisions that articulate pedagogical, institutional, and technological aspects, avoiding out-of-context approaches or those driven exclusively by operational criteria.
Cross-cuttingly, the centrality of teaching practice as a structuring element of the educational process stands out. Pedagogical mediation, didactic intentionality, and the capacity for critical interpretation of technologies are essential conditions for AI to effectively contribute to the quality of education. The preservation of this role ensures that technology is integrated into formative processes as a support resource, rather than as a replacement for the educational relationship.
In Basic Education, the use of AI must be articulated with the integral development of students, promoting learning that enables them to critically understand, use, and evaluate these technologies. The integration between the pedagogical use of AI and the development of competencies related to its understanding contributes to the formation of critical, autonomous subjects prepared for digital citizenship. The use of AI should contribute to the formation of students capable of critically understanding the impacts of technology on society, exercising their citizenship in the digital environment, and participating in an ethical, responsible, and conscious manner in social processes mediated by emerging technologies.
In Higher Education, the incorporation of AI broadens the possibilities for academic training and professional practice, while demanding the development of critical, ethical, and technical competencies. The articulation between academic knowledge, the use of technology, and reflection on its impacts is fundamental for training professionals capable of operating in complex and constantly transforming contexts.
This Opinion establishes limits and safeguards for the use of AI in education, especially regarding the protection of personal data, transparency, human supervision, and the preservation of teaching activities and institutional responsibility. It also recognizes that different applications present distinct levels of impact on pedagogical, evaluative, and management processes, which is why it adopts an approach guided by regulatory proportionality and risk classification. This perspective allows governance, monitoring, and rights-protection mechanisms to be compatible with the nature, purpose, and complexity of each context of use, reconciling innovation, security, flexibility, and respect for educational purposes.
In this context, the growing incorporation of AI-based systems reinforces the need for permanent mechanisms of transparency, traceability, human oversight, and bias mitigation, as conditions for responsible governance of these technologies. Social trust in their educational use depends not only on their technical efficacy, but on the institutions' capacity to ensure that their adoption occurs in an ethical, verifiable, proportional manner and committed to the protection of fundamental rights, the integrity of educational processes, and the full development of students.
At the same time, the importance of the autonomy of education systems and educational institutions in defining their implementation strategies is recognized, respecting their specificities and contexts. This autonomy, however, must be exercised in accordance with established principles and guidelines, ensuring coherence and responsibility in the use of technology.
The effectiveness of the guidelines established in this Opinion will depend on the capacity of educational systems and institutions to promote permanent processes of training, professional development, institutional strengthening, and the improvement of their pedagogical and governance practices. In a context of rapid technological evolution, the continuous updating of knowledge and skills is an essential condition for education professionals to act as critical, ethical, and qualified mediators in the use of AI, promoting safe, responsible, and pedagogically oriented learning environments.
It is also recognized that the implementation of these guidelines must consider the different Brazilian educational realities, observing the infrastructure conditions, connectivity, professional training, and institutional capacities existing in the various education systems and networks. Strengthening the national capacity for scientific, technological, and educational production related to AI, combined with cooperation between institutions, constitutes a strategic element to expand the country's autonomy, value its cultural diversity, and ensure that technological innovation contributes to the reduction of inequalities and to the development of Brazilian education.
AI should be incorporated into educational processes as an instrument to strengthen intellectual autonomy, authorship, academic honesty, critical reflection, and the collaborative construction of knowledge, expanding human capabilities to learn, create, investigate, and solve problems, without replacing the active participation of students and education professionals in formative processes.
Considering the continuous evolution of AI technologies, it is recommended that the National Education Council maintain a permanent mechanism to monitor the implementation of these Guidelines, in coordination with educational systems, educational institutions, the scientific community, experts, and representatives of society, with a view to monitoring their application, identifying emerging regulatory gaps, and proposing potential normative updates. It is recommended that this review take place, preferably, in biennial cycles or whenever relevant changes in the technological, scientific, pedagogical, or regulatory scenario so justify.
It is recorded that the present Opinion, once approved by the National Education Council, will constitute the legal, pedagogical, and institutional basis for the elaboration of the corresponding Resolution, intended to regulate the implementation of the Guidelines hereby approved and give them normative effectiveness. To this end, the Bicameral Commission will continue its work with a view to elaborating the draft Resolution, observing the principles, guidelines, limits, safeguards, and parameters established in this Opinion, which shall guide the drafting and interpretation of the subsequent normative act.
**II. VOTES OF THE RAPPORTEURS**
We vote for the approval of the Guidelines for the Use of Artificial Intelligence - AI in Brazilian Education, in the form of this Opinion and the attached Draft Resolution, of which it is an integral part.
Brasília-DF, September 1, 2026.
Councilor Celso Niskier - Rapporteur
Councilor Israel Matos Batista - Rapporteur
**III. DECISION OF THE PLENARY COUNCIL**
The Plenary Council unanimously approves the Rapporteurs' vote.
Session Room, September 1, 2026.
Councilor Antonio Cesar Russi Callegari is President
**REFERENCES**
BRAZIL. Constitution (1988). Constitution of the Federative Republic of Brazil. Brasília, DF: Federal Senate, 1988.
BRAZIL. Law No. 9,394, of December 20, 1996. Establishes the guidelines and bases of national education. Official Gazette of the Union: Brasília, DF, Dec. 23, 1996.
BRAZIL. Law No. 13,005, of June 25, 2014. Approves the National Education Plan (PNE). Official Gazette of the Union: Brasília, DF, June 26, 2014.
BRAZIL. Ministry of Education. National Common Curricular Base. Brasília, DF: MEC, 2017.
BRAZIL. Law No. 13,709, of August 14, 2018. General Personal Data Protection Law (LGPD). Official Gazette of the Union: Brasília, DF, Aug. 15, 2018.
BRAZIL. Ministry of Science, Technology and Innovations. Brazilian Artificial Intelligence Strategy. Brasília, DF: MCTI, 2021.
BRAZIL. Ministry of Education. Artificial intelligence in Basic Education: guiding document on curricular paths and ethical practices for the use of AI in schools. Brasília, DF: MEC, 2026.
BRAZIL. Ministry of Education. Framework for the Responsible Development and Use of Artificial Intelligence in Education. Brasília, DF: MEC, 2024.
BRAZIL. National Education Council. Resolution CNE/CEB No. 2, of March 21, 2025. Guidelines for Digital and Media Education. Brasília, DF: CNE, 2025.
SCIENTIFIC COMMITTEE OF THE SCIENCE FOR CHILDREN CENTER. Study No. 15: Protection in early childhood between screens and digital media. Science for Children Center, 2025. Available at: www.ncpi.org.br Accessed on: Sep 04, 2026.
OECD. Education Policy Outlook 2024: Reshaping teaching into a thriving profession from ABCS to AI. Paris: OECD Publishing, 2024.
UNESCO. Recommendations on the Ethics of Artificial Intelligence. Paris: UNESCO, 2022.
UNESCO. Guide to Generative AI in Education and Research. Paris: UNESCO, 2023.
**MINISTRY OF EDUCATION**
**NATIONAL EDUCATION COUNCIL**
**FULL COUNCIL**
**DRAFT RESOLUTION**
Establishes the Guidelines for the Use of Artificial Intelligence - AI in Brazilian Education, at all levels, stages, and modalities of teaching, and provides other measures.
**THE PRESIDENT OF THE NATIONAL EDUCATION COUNCIL** , in the exercise of the attributions conferred upon it by art. 7, § 1, items 'b' and 'f', and by art. 9, § 1, item 'g', and § 2, item 'h', of Law No. 4,024, of December 20, 1961, as amended by Law No. 9,131, of November 24, 1995, in view of the provisions of arts. 205, 206, 214, and 227 of the 1988 Federal Constitution; Law No. 8,069, of July 13, 1990 (Child and Adolescent Statute – ECA); Law No. 9,394, of December 20, 1996 (Law of Guidelines and Bases of National Education – LDB); Law No. 13,709, of August 14, 2018 (General Personal Data Protection Law – LGPD); Law No. 14,533, of January 11, 2023 (National Digital Education Policy – PNED); Law No. 15,211, of September 17, 2025 (Digital Child and Adolescent Statute – Digital ECA); Law No. 15,388, of April 14, 2026 (National Education Plan – PNE); and CNE/CEB Resolution No. 2, of March 21, 2025, and based on CNE/CP Opinion No. 15/2026, approved by Dispatch of the Minister of State for Education, published in the Official Gazette of the Union on xx of xx of 20xx, resolves:
**CHAPTER I**
**GENERAL PROVISIONS**
Art. 1 Guidelines for the Use of Artificial Intelligence IA in Brazilian Education are hereby established, to be observed by educational systems and educational institutions, public and private, at all levels, stages, and modalities of education.
1 These Guidelines establish parameters so that the incorporation of AI in educational systems and educational institutions occurs in an ethical, responsible, proportional, and pedagogically justified manner.
2 The purpose of these Guidelines is to ensure that each application is based on a legitimate educational purpose, necessity, proportionality, human supervision, and the protection of fundamental rights, preserving the centrality of pedagogical mediation and the formative purposes of education.
3 The use of AI in educational processes must occur in a manner compatible with the objectives of education, the protection of students, and the principles established in this Resolution.
§ 4º For the purposes of implementing these Guidelines, a distinction is made between the individual use of AI by students and teachers in diverse environments with limited institutional control capacity, and the use promoted by educational institutions, guided by pedagogical objectives, teacher mediation, data protection, governance mechanisms, and training for the responsible use of technology.
Art. 2º For the purposes of this Resolution, the following are considered:
I - general-purpose AI system: an application developed without a specific educational purpose, even if capable of being used in educational contexts;
II - specific educational solution: an AI application conceived or adapted specifically for educational purposes;
III - meaningful human oversight: human action with the effective capacity to understand, monitor, interpret, challenge, review, correct, or interrupt the results produced by AI systems, exercised under appropriate institutional conditions.
IV - exclusively automated decision: a decision produced by an AI system without effective, prior, and qualified human validation; and
V - profiling: automated processing of personal data intended to evaluate, classify, predict, or infer aspects related to the behavior, performance, characteristics, preferences, interests, risks, trajectories, or vulnerability of students or other members of the educational community.
§ 1º The AI referred to in this Resolution encompasses, among other applications, predictive systems, generative systems, intelligent agents, and digital tutors, with different purposes, degrees of autonomy, levels of transparency, auditability, and risk profiles.
2 This Resolution is integrated, in the form of the Annexes, by the Glossary of Words, Terms and Expressions and the Interpretation Guide for the Risk Classification used in the context of Artificial Intelligence in Brazilian education, which constitute a conceptual reference for the interpretation and application of these Guidelines.
Art. 3 The use of Artificial Intelligence systems, whether of general purpose or designed for educational purposes, shall always be analyzed in light of the pedagogical purpose, regulatory proportionality, human supervision, personal data protection, and potential impacts on educational processes and students' rights.
**CHAPTER II**
**ON THE PRINCIPLES**
Art. 4 The use of Artificial Intelligence in Brazilian education shall be guided by the following principles:
I - centrality of human steering, with active pedagogical mediation;
II - curricular integration of Artificial Intelligence teaching, with progressivity and pedagogical intentionality;
III - personal data protection, digital ethics, and information security;
IV - inclusion, equity, accessibility, and digital citizenship;
V - appreciation of epistemic, cultural, and social diversity;
VI - pedagogical autonomy of educational institutions;
VII - scientific sovereignty and the development of national solutions;
VIII - algorithmic transparency;
IX - social participation and control;
X - digital sustainability and environmental responsibility;
XI - digital well-being of students, teachers, and school administrators; and
XII - federative cooperation, mutual technical support, and the strengthening of the capacities of educational networks and systems.
§ 1º The centrality of human governance, referred to in item I of *caput* , comprises:
I - the responsibility for the conception, guidance, monitoring, and evaluation of educational processes;
II - the planning of activities, the definition of learning objectives, the selection of technologies, the monitoring of educational processes, and the critical evaluation of the results achieved; and
III - the prohibition of replacing teaching activities or institutional responsibility with automated systems.
§ 2º The centrality of human direction and pedagogical mediation does not imply the individualized transfer to education professionals of technical, legal, administrative, contractual, or security responsibilities that fall under the purview of educational systems, educational institutions, sponsoring bodies, managers, technical teams, and providers of technological solutions.
§ 3º Teaching activity must be ensured under appropriate institutional conditions, with training, technical support, working time, clear guidelines, pedagogical resources, and governance mechanisms that allow for the qualified exercise of human mediation.
Art. 5º The contracting instruments for AI systems for use in education must ensure governance, auditing, monitoring, and accountability mechanisms directed at suppliers and other agents responsible for the development, provision, or operation of AI solutions.
**CHAPTER III**
**ON THE ADOPTION, USE, AND GOVERNANCE OF ARTIFICIAL INTELLIGENCE**
Art. 6º The adoption of AI applications by educational systems and educational institutions shall be preceded by a substantiated technical, pedagogical, and institutional decision, with the participation of the educational community, considering, among others, the following criteria:
I - the pedagogical consistency of the adopted solutions and the existence of evidence of their effectiveness in supporting teaching and learning processes;
II - the promotion of well-being and the integral development of students, considering cognitive, social, and emotional dimensions;
III - the preservation of the autonomy and integrity of teaching work, avoiding the replacement of their performance;
IV - the guarantee of security, equity, transparency, and accountability in the use of technologies and data;
V - the analysis of the governance, storage, processing, and data protection conditions of the educational data involved, complying with current legislation and the protection of the strategic interests of Brazilian education;
VI - the verification of the institutional capacity for the pedagogical use of Artificial Intelligence, including the qualification of the professionals involved in its application;
VII - the evaluation of the technological infrastructure, connectivity, technical support, and operational sustainability conditions necessary for the implementation of the adopted solutions;
VIII - the analysis of the potential impacts of technology adoption on educational inclusion, considering the different student profiles, institutional contexts, and regional realities;
IX - the alignment of the solution with Brazilian linguistic, cultural, pedagogical, and administrative contexts and with the profile of students served by the network or institution; and
X - the potential of the solution for the elimination of barriers to learning and for the promotion of accessibility, in terms of Law No. 13.146, of July 6, 2015.
Sole Paragraph. When prior evidence of efficacy is insufficient, adoption shall be accompanied by subsequent institutional evaluation of its results, the findings of which shall subsidize the maintenance, revision, or interruption of the application's use.
Art. 7 For the purposes of adoption, governance, monitoring, and institutional evaluation, AI applications shall be understood, according to their predominant purposes of use, distinguishing, among others, applications aimed at:
I - pedagogical support;
II - institutional and administrative management;
III - evaluation processes; and
IV - accessibility and inclusion actions.
§ 1º When a given application presents characteristics compatible with more than one purpose or risk category provided for in this Resolution, the applicable criteria shall be considered cumulatively, and the classification corresponding to the highest risk level shall prevail for the purposes of defining safeguards, supervision mechanisms, and institutional governance.
§ 2º The classification shall consider the predominant purpose of use, the potential impact on students' rights, the degree of autonomy of the application, the nature of the processed data, and the relevance of the educational decisions supported or influenced by the AI system.
Art. 8º AI may be used as a support resource for pedagogical practices, in a complementary character to teaching activities, for the diversification of teaching strategies, the monitoring of learning, the personalization of educational experiences, the promotion of accessibility, and the organization of educational activities.
§ 1º The effectiveness of the pedagogical use of AI presupposes clear pedagogical intentionality that guides the selection, use, and evaluation of resources, with teacher mediation being responsible for ensuring the contextualization of contents, the critical interpretation of the produced information, and the adaptation of activities to the students' needs.
2 The pedagogical use of AI will consider the ethical and responsible management of these tools at all stages of their use, including the formulation of prompts, questions and instructions, interaction with systems, critical analysis of the responses produced and the use of generated content.
Art. 9 In Distance Education, the use of AI systems may support pedagogical mediation activities, tutoring, guidance, learning monitoring, student support and personalization of educational pathways, provided it is subordinated to institutional pedagogical planning, teacher curation and meaningful human supervision.
§ 1º The use of these technologies shall not characterize the replacement of the academic responsibilities proper to the teacher, the pedagogical mediator, the tutor, or the educational institution, nor the suppression of the principles of collective, interactive, and relational construction of formative pathways, and must contribute to expanding pedagogical interaction, accessibility, monitoring, and the quality of formative processes.
§ 2º The mediation performed by AI systems shall not be computed for the purpose of fulfilling the regulatory requirements of tutoring and interaction provided for in Distance Education regulations, and its use must be subject to verification in the evaluation, regulation, and supervision instruments of distance Higher Education.
Art. 10. The use of AI tools in academic and educational activities shall be guided by the values of intellectual authorship, academic honesty, and responsibility for the content produced.
§ 1º Interaction with AI systems does not transfer responsibility for the presented content to the technology; the student, teacher, or researcher bears the intellectual responsibility for the selection, verification, truthfulness, consistency, and adequacy of the information used or produced, as well as transparency regarding the use of AI and the attribution of authorship compatible with effective human contribution.
§ 2º The protection of the image, voice, personal data, and identity of education professionals and students is guaranteed against the use of AI systems.
Art. 11. Educational systems and institutions shall promote ongoing training and updating programs for education professionals regarding the pedagogical use of Artificial Intelligence, addressing, among other aspects:
I - algorithmic literacy;
II - drafting of prompts and instructions for AI systems;
III - critical curation of generated content.
IV - assessment mediated by digital technologies;
V - personal data protection;
VI - academic integrity; and
VII - identification and mitigation of algorithmic biases.
Art. 12. The use of AI in the educational context requires the adoption of institutional governance mechanisms that ensure transparency, control, and continuous monitoring, allowing the identification of the purposes of the applications used, the analysis of the risks involved, and the definition of criteria for their evaluation and review.
§ 1. Institutional documentation will record decisions related to the adoption and use of AI, as well as the parameters that guide its use.
§ 2. Institutional governance processes will include mechanisms for periodic review of policies, procedures, and criteria related to the use of AI, considering technological changes, regulatory updates, new scientific evidence, and the results observed in its implementation.
3 Faculty participation will be present in the monitoring and supervision processes, ensuring that the use of technology remains aligned with pedagogical purposes and that any adjustments are made in a well-founded manner.
4 When human review involves education professionals, educational systems and institutions will ensure adequate conditions for its exercise, including specific training, institutional time, technical support, consultation channels, and decision-making support.
Art. 13. The use of AI in educational processes will be clearly communicated to students, teachers, managers, and, where applicable, to families or guardians, allowing knowledge about its application and objectives.
Sole paragraph. The results generated by AI systems will remain subject to critical analysis and human responsibility, in order to ensure their pedagogical adequacy, reliability, and compliance with the principles of this Resolution, as well as to prevent the reproduction of errors, biases, or inadequate information.
Art. 14. The replacement of teaching performance by automated systems, as well as the transfer of pedagogical responsibility to AI systems, is prohibited.
§ 1. Processes related to learning assessment, academic monitoring, educational guidance, and decision-making with repercussions on students' trajectories will remain subject to human supervision and validation compatible with their educational relevance.
2 The use of AI as a resource to support the creation, adaptation, and review of didactic materials and pedagogical activities must be clearly informed and does not remove the author's responsibility for verifying and adapting the content to the learning objectives.
Art. 15. The use of content produced by teachers and students for training third-party AI models is prohibited, unless specifically authorized.
Sole Paragraph. The exception provided for in the caput does not apply to content produced by children and adolescents, whose use for training third-party AI models is prohibited.
Art. 16. The use of AI systems for targeted advertising, marketing profiling, or commercial exploitation of educational data of students and teachers is prohibited in educational systems and institutions.
**CHAPTER IV**
**ON THE USE OF ARTIFICIAL INTELLIGENCE IN BASIC EDUCATION**
Art. 17. In Basic Education, the use of AI shall observe the principles established in art. 4 and, especially, the integral protection of children and adolescents, age appropriateness, the progressivity of learning, and the integral development of students, respecting the different stages of cognitive, emotional, social, and ethical development and preserving the centrality of human interactions in educational processes.
Art. 18. The use of AI in Basic Education shall observe the following guidelines:
I - the promotion of the integral development of students, considering the cognitive, social, and emotional dimensions;
II - the use of resources that promote educational inclusion, accessibility, and the personalization of learning experiences, respecting the characteristics and needs of students;
III - the preservation of students' intellectual autonomy and well-being, avoiding excessive dependence on technologies;
IV - the integration of AI literacy in a manner coordinated with digital, media, and computational literacies;
V - the consideration of the digital access conditions of students and schools, with the adoption of pedagogical alternatives that avoid exclusion resulting from limitations in infrastructure, connectivity, or device availability;
VI - compliance with the age appropriateness of tools, contents, and educational experiences supported by AI;
VII - the strengthening of students' ability to critically analyze information, synthetic content, and recommendations produced by automated systems;
VIII - the promotion of safe, respectful digital environments compatible with the rights of children and adolescents;
IX - the strengthening of the critical and ethical mediation capacity of education professionals; and
X - the prevention of inadequate, invasive, or disproportionate monitoring practices of students.
§ 1 The adoption of AI systems in Basic Education shall be accompanied by teaching about AI, in order to promote intentional and meaningful learning about these systems in school curricula.
§ 2 Age appropriateness is not limited to defining age-based restrictions on access to specific technologies, but shall also cover the conditions for technology use, especially regarding pedagogical mediation, human supervision, personal data protection, and the development of skills for critical, ethical, and responsible use of these technologies.
3 The distinction between direct, autonomous, individual, or unsupervised student access and mediated or supervised access by education professionals does not exempt, under any circumstances, the pedagogical monitoring by systems, schools, and education professionals of the use of AI in schools.
Art. 19. The integration of AI into the Basic Education curriculum shall take place in articulation with digital, media, and computational education, observing the National Common Curricular Base (BNCC), the National Operational Guidelines on the use of digital devices in school spaces and the curricular integration of digital and media education, the National Curricular Guidelines for Early Childhood Education, Elementary Education, High School, Integral Education, and the different modalities of Basic Education, as well as other current national guidelines, in a cross-cutting, progressive manner appropriate to the different stages, years, modalities, institutional contexts, and student development characteristics.
§ 1º AI may be treated both as an object of learning and as a pedagogical resource, provided that its use is aligned with educational objectives and pedagogical practices.
2 The curricular integration of AI will consider the social, cultural, and technological contexts of students and institutions and will favor articulation between different areas of knowledge, observing the commitment to reducing educational inequalities.
Art. 20. Basic Education will promote the development of learning that allows students to understand the fundamentals, uses, and impacts of AI, contemplating, among other aspects:
I - the basic understanding of the functioning of computer systems and AI applications, including their modes of operation, possibilities, limitations, and impacts on contemporary society;
II - the understanding of the role of data and its utilization in AI systems;
III - the development of logical, analytical, and critical thinking;
IV - the identification of limits, risks, and possible biases present in technologies;
V - understanding the risks associated with misinformation, digital manipulation, synthetic content, and automated influence strategies present in digital environments;
VI - developing the ability to verify sources, validate information, and exercise critical judgment regarding content produced or mediated by AI;
VII - understanding the difference between technological support and independent intellectual creation, with the development of attitudes consistent with academic honesty;
VIII - critical analysis of the social, economic, cultural, political, environmental, and ethical impacts associated with the development and use of AI;
IX - the promotion of responsible, ethical, and safe behaviors in the digital environment; and
X - the development of metacognitive capabilities related to the use of AI, encouraging students to reflect critically on their own processes of thought, learning, decision-making, and knowledge production mediated by digital technologies.
Single Paragraph. In the stages and modalities of Basic Education where the use of AI tools or applications by students is permitted, this contact with concepts, tools, and applications related to AI shall occur progressively, mediated, contextualized, and appropriate to the students' development, pedagogical purposes, curricular specificities, and institutional conditions of each stage and modality.
Art. 21. In Early Childhood Education, AI systems may be used by education professionals in planning and pedagogical management activities and as accessibility resources aimed at the elimination of barriers.
§ 1º The use of any AI systems by students is prohibited, whether directly, autonomously, and individually by the child, even if mediated by education professionals.
§ 2º The use of accessibility resources shall observe the exclusive purpose of eliminating barriers, provided they do not imply sensitive inferences, individualized profiling, or decisive repercussions on the student.
Art. 22. In the Early Years of Elementary Education, any eventual use of AI-based resources shall occur exclusively through pedagogical activities planned, conducted, and mediated by education professionals.
§ 1º Direct, autonomous, individual, or unsupervised access by students to AI systems is prohibited.
§ 2º Student interaction with generative or conversational AI systems is prohibited, even under the supervision of education professionals.
§ 3º The incorporation of these technologies shall occur gradually, with progressive expansion of their use as digital literacy, critical thinking, digital citizenship, and intellectual autonomy skills are developed.
§ 4º Whenever the institutional use of AI applications involves students in the Early Years of Elementary School, educational systems and institutions shall ensure transparency regarding the pedagogical purposes of the technology employed, promote clear communication to families or guardians, and strictly observe the protection of personal data.
§ 5º The provisions of this article do not apply to assistive technologies and accessibility resources intended for the elimination of learning barriers, observing the provisions of art. 36, § 3º.
Art. 23. In the Final Years of Elementary School, the use of generative AI tools may occur only in pedagogical activities mediated by education professionals.
Sole paragraph. The direct, autonomous, individual, or unsupervised use of these tools by students is prohibited.
Art. 24. At any stage of Basic Education, the use of AI systems for targeted advertising, market profiling, or commercial exploitation of student data is prohibited.
**CHAPTER V**
**ON THE USE OF ARTIFICIAL INTELLIGENCE IN HIGHER EDUCATION**
Art. 25. In Higher Education, AI may be integrated into academic training both as an object of study and as a support resource for teaching, research, and outreach activities, in a planned and coordinated manner, promoting the connection between theoretical foundations, practical applications, and critical reflection, respecting the specificities of the courses, the National Curriculum Guidelines, and the institutional pedagogical projects.
Art. 26. In teacher education degree courses, training related to AI shall explicitly cover teacher digital literacy, the pedagogical use of emerging technologies, the critical analysis of educational data, technology-assisted assessment, and the ethical aspects related to the use of AI in school environments.
Sole paragraph. Knowledge related to AI shall be articulated, whenever possible, with supervised internship, extension, and research activities, favoring the approximation between teacher training and the concrete challenges faced by Basic Education institutions.
Art. 27. Higher education institutions shall establish their own policies and guidelines for the use of AI in teaching, research, extension, and academic management activities, encompassing principles of authorship, academic integrity, transparency, intellectual responsibility, and rights protection.
§ 1º These policies shall, whenever relevant, be incorporated into the Institutional Development Plan – IDP, the Pedagogical Course Projects – PCP, bylaws, academic regulations, and other institutional governance instruments.
§ 2º Institutional policies may contemplate criteria for transparency regarding the use of technology, forms of declaration of use when relevant, guidelines for the prevention of practices incompatible with academic honesty, and procedures for monitoring, reviewing, and continuously improving institutional practices related to AI.
§ 3º In institutions offering undergraduate teaching degrees, these policies will also contemplate initial teacher training strategies, preparing future teachers to act critically, ethically, and pedagogically qualified in contexts increasingly mediated by digital technologies and intelligent systems.
Art. 28. In the field of scientific research, the use of AI shall observe the principles of scientific integrity, methodological transparency, and intellectual responsibility, and must be declared whenever its contribution is relevant to the drafting of texts, data analysis, bibliographic review, programming, or other stages of the investigative process.
Sole paragraph. The use of these technologies does not transfer authorship or responsibility for the produced results to the AI, which remain entirely attributed to the researchers.
Art. 29. AI may be used in academic and school management, preserving meaningful human supervision and ensuring the possibility of reviewing decisions.
Sole paragraph. The adoption of exclusively automated decisions with a relevant impact on students' rights or academic trajectories is prohibited.
Art. 30. Higher Education shall promote the development of competencies that enable students to use AI in a critical, ethical, and responsible manner, including, among others:
I - the ability to use AI as a support tool for decision-making in different professional contexts;
II - the development of skills for data analysis, interpretation, and use;
III - the critical understanding of the possibilities, limits, and risks associated with the use of AI in learning assessment processes, academic monitoring, and educational evidence analysis;
IV - the understanding of the ethical, legal, and social implications related to the use of technology;
V - the comprehension of the conceptual foundations of AI and its impacts on the processes of knowledge production, work organization, exercise of citizenship, and social transformation;
VI - the development of the capacity for critical and metacognitive reflection on the processes of knowledge production, learning, and decision-making mediated by AI;
VII - the responsible use of AI, respecting the principles of academic integrity and professional ethics, with an understanding of the limits between technological support and one's own intellectual production; and
VIII - the recognition of the importance of proper authorship attribution and responsibility for the content produced or presented with the support of AI.
**CHAPTER VI**
**ON THE USE OF ARTIFICIAL INTELLIGENCE IN PROFESSIONAL AND TECHNOLOGICAL EDUCATION**
Art. 31. The incorporation of AI into Professional and Technological Education must contribute to the strengthening of the culture of innovation, applied research, creativity, lifelong learning, and the ability to adapt to technological transformations, in line with the needs of sustainable development, national competitiveness, and the promotion of decent work.
Art. 32. In Professional and Technological Education, the use of AI must contribute to the integral formation of students, promoting the development of technical, digital, ethical, and socio-emotional skills necessary for critical, responsible, and innovative performance in professional contexts increasingly mediated by digital technologies.
Art. 33. In courses involving laboratory activities, professional practices, internships, simulations, or the use of specialized equipment and environments, AI may constitute a support instrument for the training process, provided that its use does not compromise the development of practical competencies, professional autonomy, technical judgment, activity safety, and ethical responsibility inherent to the exercise of the professions.
Art. 34. Higher Technology Courses, in the use of AI, shall observe the provisions of art. 32 of this Resolution, without prejudice to other applicable provisions to Higher Education set forth in Chapter V.
Art. 35. High school technical courses shall promote, in their curricular organization, the development of competencies that allow the use of AI in a critical, ethical, and responsible manner, contemplating:
I - the development of skills for data analysis, interpretation, and use;
II - the ability to use AI as a support tool in different professional contexts;
III - the critical understanding of the possibilities and risks associated with the use of AI in learning, assessment, and analytical processes, as well as in productive processes in the world of work; and
IV - the responsible use of AI-mediated tools, including the understanding of the risks associated with disinformation, digital manipulation, and automated influence strategies present in digital environments.
**CHAPTER VII**
**ON RISK CLASSIFICATION, SAFEGUARDS, AND REGULATORY PROPORTIONALITY**
Art. 36. Educational systems and educational institutions shall adopt an approach guided by the principle of regulatory proportionality, according to which the levels of supervision, governance, documentation, transparency, and monitoring shall be compatible with the educational purpose, the system's degree of autonomy, and the risks effectively associated with the use of AI applications.
1 Risk classification will consider, in an integrated manner, the educational purpose of the application, the degree of system autonomy, the type of data processed, the potential impact on students' rights, the level of decision automation, the possibility of profiling, and the reversibility of the effects produced.
§ 2º Risk analysis will not be restricted to the technical characteristics of the solutions employed, also considering the concrete conditions of their use, including the existence of human supervision, pedagogical mediation, governance mechanisms, data protection, and institutional monitoring capacity.
§ 3º Restrictions related to the use of profiling mechanisms do not preclude the adoption of assistive technologies and resources intended exclusively for promoting accessibility, educational inclusion, and the elimination of learning barriers, provided that their use complies with the principles of data protection, human supervision, proportionality, and respect for students' rights.
Art. 37. For the purposes of organizing governance, monitoring, and risk mitigation practices, AI applications used by educational systems and institutions shall be classified into the following risk levels:
I - low-risk applications;
II - moderate risk applications;
III - high-risk applications; or
IV - applications of excessive risk or incompatible with educational purposes.
Sole Paragraph. The classification into risk levels provided for in this article establishes the corresponding categories and minimum safeguards, and the list of applications indicated in this Resolution is illustrative and not exhaustive.
Art. 38. Low-risk applications are considered to be those for instrumental, organizational, complementary pedagogical or accessibility support which, considering the purpose, context of use, nature of the processed data and the degree of autonomy of the system, do not produce a relevant impact on students' rights, evaluation processes, academic decisions, educational trajectories or conditions of access, permanence, progression and certification, nor involve individualized profiling, processing of sensitive personal data, continuous monitoring, continuous interaction with students or relevant automated decision-making.
§ 1º They are low-risk applications, including:
I - the organization, systematization, or adaptation of teaching materials;
II - auxiliary tools for pedagogical or administrative planning;
III - non-evaluative textual revision;
IV - support for translation, summarization, formatting, or content adaptation; and
V - simple accessibility resources aimed at eliminating barriers, such as transcription, subtitling, screen reading, text enlargement, or format conversion, provided they do not imply sensitive inferences, individualized profiling, or decisive repercussions on the student.
§ 2º In the hypotheses dealt with in this article, the following will be observed, as parameters of good institutional governance: duties of transparency, information security, personal data protection, institutional accountability, simplified record of the purpose of use, and possibility of human supervision whenever the application is used in institutional educational activity.
Art. 39. Applications used for pedagogical, academic, administrative, or institutional support are considered moderate risk. Although they do not produce relevant automated decisions, they involve continuous interaction with students, generation of academic recommendations, personalization of learning paths, support for intellectual production, formative monitoring, institutional guidance, or systematic treatment of educational information, provided they do not produce automatic effects on grades, certification, progression, retention, benefit granting, disciplinary sanctions, selection, or other sensitive educational decisions.
§ 1º Moderate risk applications include, among others:
I - automated tutoring or pedagogical support systems;
II - formative feedback tools without direct academic repercussions;
III - institutional assistants used for general academic support, guidance, or informational assistance;
IV - support tools for text production, research, or study organization, when used in a pedagogical or academic context;
V - academic recommendation mechanisms without automatic decision-making effect;
VI - learning monitoring systems without direct and automatic repercussions on grades, certification, permanence, benefits, or sanctions; and
VII - accessibility resources involving personalization, individualized adaptation, or systematic processing of educational information, provided they do not involve sensitive data or relevant automated decisions.
§ 2º In the hypotheses addressed in this article, educational systems and institutions shall adopt safeguards proportional to the context of use, especially:
I - clear communication to the educational community about the use of technology, its purposes, its limits, and institutional responsibilities;
II - institutional record of the applications used, indicating purpose, audience reached, data processed, and those responsible for monitoring;
III - mandatory human review of results whenever there is an influence on pedagogical guidance, academic monitoring, personalized learning paths, or recommendations directed at the student;
IV - periodic monitoring of errors, biases, disproportionate impacts, and pedagogical adequacy;
V - prohibition of the use of educational data for training third-party models, except for specific and revocable authorization, with a legitimate purpose, transparency, and compatible safeguards, observing the prohibition referred to in art. 41, item XI;
VI - provision of an institutional channel for clarification, review, or contestation when the recommendation produced may affect the student's educational experience; and
VII - periodic evaluation of the adequacy of the application to the originally assigned risk level, especially when there is an expansion of functionalities, integration with institutional databases, or change of purpose.
Art. 40. Applications capable of producing, supporting, or influencing relevant effects on students' rights, academic life, evaluation processes, certification, selection, permanence, progression, retention, granting of benefits, disciplinary analysis, distribution of educational opportunities, or other sensitive educational processes are considered high risk.
§ 1º High-risk applications include, among others:
I - automated support systems for correcting objective or closed-ended assessments with academic repercussion, which may only be used as an auxiliary instrument, prohibiting the automatic assignment of a grade, concept, approval, disapproval, certification, progression, retention, or any academic consequence without effective, prior, qualified, and documented human validation;
II - the systems of *proctoring* or other monitoring technologies used in assessment processes;
III - individualized academic profiling;
IV - systems related to selection, certification, granting of benefits, permanence, progression, retention, or disciplinary analysis;
V - applications involving the processing of sensitive personal data, including data concerning disability, health, biometrics, behavior, social vulnerability, or sensitive individual performance;
VI - applications involving relevant behavioral inferences, identification of individualized educational risk patterns, or prediction of dropout, performance, vulnerability, or conduct;
VII - applications that may produce a relevant impact on educational opportunities, access to institutional resources, or conditions for continuing studies;
VIII - tools for *feedback* , tutoring, recommendation, or monitoring that come to influence grades, progression, certification, retention, benefits, sanctions, or other relevant institutional decisions;
IX - automated tools for detecting AI-generated content, when used in investigations with potential academic or disciplinary repercussion, admitted exclusively as an auxiliary element of analysis; and
X - automated pre-correction systems for essay-type assessments, essays, authored textual productions, and other instruments that require interpretive, evaluative, or argumentative judgment regarding the student's intellectual production, admitted exclusively as a preparatory and auxiliary instrument for the correction performed by the teacher, observing the provisions of § 4º.
§ 2º In the hypotheses dealt with in this article, enhanced safeguards compatible with the sensitive nature of the application shall be ensured, including:
I - prior algorithmic impact assessment, proportional to the risk and educational purpose;
II - personal data protection impact report, when applicable; and
III - significant, continuous, qualified, and documented human supervision.
IV - mechanisms for institutional traceability, documentation, and auditing;
V - transparency and accessible explanation regarding the purpose, general operation, limits, and potential impacts of the application;
VI - effective possibility of contestation, human review, and correction of results, recommendations, or decisions supported by Artificial Intelligence systems;
VII - prior and documented human validation of any result with evaluative, disciplinary, academic, administrative, or certificating repercussion;
VIII - periodic monitoring of biases, systematic errors, and discriminatory or disproportionate impacts on specific groups;
IX - adoption of less invasive alternatives whenever available, especially in evaluative, monitoring, individualized tracking, or sensitive data processing procedures; and
X - guarantee that the application does not replace the responsibility of an education professional or competent institutional body.
§ 3º The use of high-risk applications may not result in an exclusively automated decision with a relevant effect on the student's educational trajectory, with the assignment of grades, certification, progression, retention, selection, granting of benefits, application of sanctions, and other relevant academic or institutional decisions remaining under the responsibility of an education professional or competent institutional body, ensuring effective human supervision, justification, recording, and the possibility of review.
§ 4º For the purposes of item X of § 1º, pre-correction is considered to be the preparatory stage, devoid of its own evaluative effect, in which the AI system can assist human correction, complying cumulatively with:
I - the prohibition that the score, grade, or merit generated by an automated system produce academic effect without human correction; and
II - the guarantee of complete reading and direct appreciation of the student's production, as well as the right to review by a human evaluator, prohibiting tacit approval, batch validation, and mere ratification of the result suggested by the system.
Art. 41. The following are incompatible with educational purposes and the comprehensive protection of students, and their use in educational systems and educational institutions is prohibited, among others:
I - social scoring systems;
II - emotional surveillance and the recognition or inference of emotions of students and education professionals;
III - the exploitation of student vulnerabilities, including age-related, emotional, socioeconomic, cognitive, or behavioral vulnerabilities;
IV - psychological, biometric, or behavioral profiling for classificatory, disciplinary, commercial, or educational opportunity restriction purposes.
V - exclusively automated decisions with relevant effects on promotion, retention, certification, permanence, granting of benefits, dismissal, selection or sanction;
VI - continuous biometric surveillance or permanent monitoring of students;
VII - the use of educational data for targeted advertising, commercial exploitation, behavioral manipulation, or purposes incompatible with the educational relationship;
VIII - systems that produce segregation, discrimination, or exclusion of students based on automated inferences about ability, behavior, vulnerability, or performance;
IX - the assignment of grades, concepts, or merit to essay evaluations, essays, authored textual productions, and other instruments that require an interpretive, evaluative, or argumentative judgment on the student's intellectual production exclusively by an AI system, or without the human correction referred to in art. 40, § 4;
X - the application of academic or disciplinary sanctions based exclusively on the result of an automated tool for detecting AI-generated content; and
XI - the use of educational data from children and adolescents to train third-party AI models.
§ 1º The potential use of ad-hoc identity verification mechanisms, when strictly necessary for access to virtual environments, assessments, or academic activities, shall not be confused with continuous biometric surveillance, emotion recognition, behavioral profiling, or permanent monitoring, and must observe the principles of purpose limitation, necessity, proportionality, personal data protection, transparency, human oversight, and the adoption of the least intrusive alternative available.
§ 2º The prohibition set forth in item II of the caput does not cover the use of assistive technology for the exclusive purpose of accessibility, with the consent of the student or their guardians, nor scientific research approved by a research ethics committee, with any evaluative, disciplinary, or selective repercussions prohibited in both cases.
Art. 42. Educational systems and educational institutions shall periodically reassess the classification of the applications used, as well as the adequacy of the safeguards and supervision mechanisms adopted, in light of the criteria established in this Resolution, in order to ensure their adherence to technological transformations, new contexts of use, and the protection of students' rights.
**CHAPTER VIII**
**FINAL AND TRANSITIONAL PROVISIONS**
Art. 43. Educational policies shall focus primarily on the qualification of institutional conditions of use, ensuring that governance requirements are compatible with the size, administrative capacity, technological infrastructure, and operating conditions of each educational system or institution, without prejudice to compliance with the principles of rights protection, transparency, human oversight, and the quality of educational processes.
Art. 44. The implementation of these Guidelines shall be guided by cooperation among the federative entities, mutual technical support, and the strengthening of the institutional capacities of education systems and networks, in compliance with the collaborative regime provided for in the Constitution, through coordinated actions for the training of education professionals, knowledge sharing, production of technical frameworks, development of institutional capacities, and dissemination of best practices.
Single Paragraph. The implementation of AI-related policies shall consider the specificities of different educational networks and institutions, observing criteria of proportionality, equity, and permanent technical support, in order to expand local capacities and prevent structural inequalities from compromising qualified access to the opportunities offered by digital technologies.
Art. 45. The National Education Council will monitor the implementation of these Guidelines, in coordination with educational systems, educational institutions, state and municipal education councils, the scientific community, specialists, and representatives of society, with a view to monitoring their application, identifying emerging regulatory gaps, and proposing potential normative updates.
Single Paragraph. The review of these Guidelines, including the Glossary in the Annex, shall occur whenever relevant changes in the technological, scientific, pedagogical, or regulatory scenario so justify.
Art. 46. Educational systems and educational institutions shall have a period of twelve months, counted from the publication of this Resolution, to promote the adaptation of their policies, practices, governance instruments, and current contracts to the provisions of these Guidelines, without prejudice to the immediate application of the prohibitions provided for in art. 41.
Art. 47. This Resolution enters into force on the date of its publication.
**ANNEX**
**GLOSSARY OF WORDS, TERMS AND EXPRESSIONS USED IN THE CONTEXT OF ARTIFICIAL INTELLIGENCE APPLIED TO EDUCATION**
Annex to the proposal for an Opinion and Resolution establishing Guidelines for the Use of Artificial Intelligence - AI in Brazilian Education (SEI Process No. 23001.001062/2024-18)
**PRESENTATION**
This Glossary is an Annex to the draft Opinion and Resolution that establish Guidelines for the Use of Artificial Intelligence - AI in Brazilian Education. Its purpose is to reduce interpretive ambiguities, standardize the use of words, terms, and expressions of a technical, pedagogical, and legal-regulatory nature, supporting the use of AI in schools and universities by all involved agents, teachers, students, managers, families, researchers, technicians, and the general community. The definitions consolidated here combine three types of sources: a) technical terminology established in national and international AI glossaries and frameworks; b) concepts extracted from the Opinion itself or required by its normative logic, such as meaningful human oversight, risk classification, teaching mediation, academic integrity, and personal data protection; and c) the legal-regulatory expressions necessary for the interpretation of normative commands, such as open clause, closed clause, illustrative list, and indeterminate legal concept. The descriptions seek clarity and precision, avoiding generic understandings incompatible with the guiding function of the document. The definitions contained in this Glossary do not replace current legislation, the 1988 Federal Constitution, the Law of Guidelines and Bases of National Education - LDB, the General Personal Data Protection Law - LGPD, the Statute of the Child and Adolescent - ECA and Digital ECA, the Civil Rights Framework for the *Internet* , the Brazilian Inclusion Law - LBI, the National Policy for Digital Education - PNED, and other applicable standards. In case of conflict or doubt, the higher-ranking standard or the interpretation most protective of the fundamental rights of students, teachers, researchers, and other subjects of the educational process shall prevail, with primacy given to integral protection and the best interests of the child and adolescent. For all purposes of this Glossary, AI must be understood as a support tool, devoid of personality, legal autonomy, or independent responsibility. Responsibility remains fully attributed to the natural and legal persons who develop, contract, make available, configure, supervise, or use the systems.
**ENTRIES**
The entries are organized in alphabetical order and numbered sequentially. Internal cross-references between entries must be interpreted systematically, in light of the Opinion and Resolution to which this Glossary is annexed.
**1. AI-powered accessibility:** Use of AI systems to reduce social barriers through communicational, sensory, cognitive, motor, or linguistic supports, by means of transcription, audio description, captioning, translation, assisted reading, interface adaptation, or equivalent resources. It is not to be confused with profiling as its exclusive purpose is inclusion.
***2*. Age appropriateness:** Principle according to which the design, selection, and use of AI systems in education must be compatible with the cognitive, emotional, social, and ethical development stage of students, especially children and adolescents. It imposes progressive restrictions on use, mandatory mediation by education professionals, and prohibition of exposure to age-incompatible features, in compliance with the Digital ECA and integral protection.
**3. AI agent:** Autonomous or semi-autonomous AI systems designed to process inputs, plan steps, trigger tools, Application Programming Interfaces - API or external services, and execute actions oriented toward previously defined specific objectives. For the educational context, their use requires explainability, transparency regarding their purpose, general programming logic, operational commands, triggered tools, and the system's operational limits, as they can produce operational, pedagogical, evaluative, administrative, or institutional effects. There must be meaningful human supervision, logging of executed actions, the possibility of interruption, review of results, and clear identification of those responsible for their configuration, authorization, and monitoring.
**4. Personal data processing agents:** Natural persons or legal entities responsible for the processing of personal data, including the controller, who makes decisions regarding processing, and the operator, who processes data on behalf of the controller. In the educational context, educational institutions ordinarily act as controllers of student, faculty, and family data, even when contracting third-party AI platforms, remaining responsible for the compliance of the processing.
**5. Hallucination:** Incorrect, fictitious, inconsistent, or factually baseless response generated by a generative AI system, even if presented with an appearance of accuracy and authority.
**6. Data Analysis:** Set of statistical, computational, or interpretive procedures intended to organize, describe, and extract information from data.
**7. Anonymization:** Technical process by which data can longer be associated, directly or indirectly, with an identified or identifiable natural person, considering reasonable and available technical means. It must be judicious, as small bases or cross-referencing with other data may allow re-identification.
**8. Machine learning:** Subfield of AI that uses data and statistical models to improve the performance of systems in specific tasks, without explicit programming for each situation. It can be used in content recommendation, predictive analysis, and recognition of educational patterns.
**9. Unsupervised learning:** Machine learning paradigm in which the model identifies patterns, groupings, or structures in unlabeled data, without predefined answers. In education, it can be used for grouping learning profiles, a hypothesis that requires caution as it approaches profiling.
**10. Reinforcement learning:** Machine learning paradigm in which the system learns by trial and error, through rewards and penalties associated with actions taken in a given environment, optimizing behavior over time. It integrates the development of adaptive systems and language models adjusted by human feedback.
**11. Deep learning:** Machine learning technique based on artificial neural networks with multiple layers, capable of learning complex data representations. It supports applications such as voice recognition, computer vision, machine translation, and large language models.
**12. Supervised learning:** Machine learning paradigm in which the model is trained from labeled data, i.e., examples accompanied by expected answers, to learn to map inputs to outputs. It is used, for example, in text classification and performance prediction. The quality and representativeness of the labels directly condition the occurrence of biases.
**13. Open architecture:** Technological solution whose code, documentation, weights, models, standards, or interfaces are made available in an open or interoperable manner, to varying degrees. It can favor auditability, local adaptation, technological sovereignty, and public control, without excluding security and governance duties.
**14. Proprietary architecture:** Model, platform, or infrastructure whose code, data, parameters, or internal functioning are controlled by a private vendor, with restricted access by users, institutions, and regulators. It demands attention to auditability, technological dependency, data protection, and sufficient transparency for educational use.
**15. AI systems audit:** Documented process of technical, legal, ethical, and pedagogical verification of AI systems, intended to evaluate compliance, security, bias, performance, traceability, data protection, and alignment with educational objectives. It must be proportional to the risk of the application.
**16. Informational self-determination:** Foundation of the personal data protection discipline that ensures natural persons control over their own information, including the right to know, access, correct, and object to treatments concerning them. In the educational environment, it guides the interpretation of consents, platform transparency, and the rights of students, families, and teachers regarding AI systems.
**17. Conventional automation:** Automatic execution of tasks previously defined by fixed rules, without relevant adaptive learning, complex inference, or autonomous content generation. It should be distinguished from AI for regulatory proportionality purposes.
**18. Intellectual autonomy:** Ability to formulate questions, build arguments, interpret data, evaluate sources, create hypotheses, sustain one's own reasoning, and exercise critical judgment.
**19. Intellectual authorship:** Assignment of intellectual responsibility to a natural person or group of persons for the conception, elaboration, argumentation, analysis, discussion, and presentation of academic or scientific work.
**20. Algorithmic Impact Assessment:** Prior and periodic procedure intended to identify, evaluate, and mitigate risks produced by algorithmic systems, including biases, errors, discrimination, opacity, pedagogical impact, exclusion, surveillance, technological dependence, and affectation of rights.
**21. Educational database:** Structured set of personal and non-personal data produced, collected, or maintained by educational institutions and teaching systems, including academic, administrative, and digital platform interaction records. Its use for the development, training, adjustment, or integration of AI systems requires a legal basis, a specific educational purpose, information security, and a prohibition of incompatible use.
**22. Digital well-being:** Condition of physical, mental and social balance in the relationship of subjects of the educational process with digital technologies, understanding the prevention of cognitive overload, hyperconnection, technological dependence and health damage resulting from the intensive use of systems, including AI.
**23. Biometrics and biometric data:** Sensitive personal data related to physical, physiological, or behavioral characteristics that allow the unique identification of a natural person, such as face, voice, iris, fingerprint, and typing patterns. Its use and processing in an educational environment require strict and justified necessity, proportionality, and the adoption of the least invasive alternative.
**24. Algorithmic opacity:** Characteristic of systems whose internal functioning is not understandable, accessible, or explainable to users, institutions, or control bodies, whether due to technical complexity or proprietary restrictions. Opacity does not eliminate the duties of transparency and explainability: when an adequate explanation is unfeasible, the use of the system in processes with a relevant impact on rights must be avoided.
**25. *Chatbot* educational:** Automated conversational system that interacts through natural language with students, families, teachers, or administrators for study support, service, guidance, or management. Its use should inform the automated nature of the interaction and preserve the human support channel when there is relevant impact or sensitive demand.
**26. Digital citizenship:** Set of knowledge, attitudes, and skills that enable the ethical, critical, safe, and responsible exercise of rights and duties in digital environments, including participation, respect, data protection, combating disinformation, and responsible use of AI.
**27. Cloud computing:** Model for providing computational resources, storage, processing, and services through remote infrastructure accessed via *the internet* . The contracting of cloud AI services by educational institutions must observe information security, data localization and legal regime, confidentiality clauses, prohibition of educational data reuse, and portability and reversibility conditions.
**28. Informed consent:** A free, informed, and unambiguous expression by which the data subject agrees to the processing of their personal data for a specific purpose. Consent cannot be presumed, unduly conditioned on access to essential activities, nor obtained in a generic manner; in the case of children and adolescents, it must observe the best interest, comprehensive protection, and, when required, the specific consent of at least one parent or legal guardian.
**29. Synthetic content:** Text, image, audio, video, or other material generated or substantially modified by AI systems.
**30. Automated grading:** Use of AI systems for grading assessment instruments.
**31. Teacher curation:** Teaching and institutional activity of selection, verification, adaptation, contextualization, and validation of content, including those generated by AI systems, in light of learning objectives, scientific accuracy, age appropriateness, and educational values.
**32. Educational data:** Data produced or processed in an educational context, including enrollment, attendance, grades, activities, platform interactions, learning records, persistence, attendance, accessibility, academic behavior, and institutional indicators, among others. They may be personal or sensitive, as applicable. They have a training purpose when used to adjust, teach, or calibrate an AI model. Their quality, representativeness, lawfulness, and diversity directly influence performance, biases, exclusions, and discriminatory risks.
**33. Sensitive personal data:** Data regarding racial or ethnic origin, religious belief, political opinion, trade union or association membership, health, sex life, genetic or biometric data, among others, when linked to a natural person. In education, they require enhanced protection, especially when relating to children and adolescents.
**34. Automated decision:** Decision produced by a computer system from data processing, with or without human participation at some stage of the process. In the educational environment, AI-supported decisions must preserve human responsibility, justification, recording, and the possibility of review.
**35. AI-generated content detectors:** Tool that estimates whether text, image, audio, or other content was produced by AI. Due to false positives and false negatives, it should not be the basis for academic or disciplinary sanctions.
**36. Grounding or *grounding* :**Process of connecting a model's response to verifiable and contextual sources, reducing hallucinations and increasing reliability. In education, it is relevant for assistants based on official documents, curricula, institutional norms, and validated materials.
**37. Prompt engineering or engineering of *prompt* :**The process of formulating, structuring, and refining prompts to guide responses from generative models. It is part of AI literacy, but does not replace conceptual domain expertise, critical judgment, authorship, or human validation.
**38. Equity in the digital context:** The principle that requires the fair distribution of opportunities for access, meaningful use, and critical appropriation of digital technologies and AI, with attention to territorial, socioeconomic, racial, gender, and disability-related inequalities.
**39. Explainability:** The ability to provide understandable explanations regarding the operation, criteria, data, limits, and outputs of an AI system.
**40. AI Governance:** The set of policies, bodies, procedures, and responsibilities for selecting, contracting, documenting, monitoring, auditing, reviewing, and discontinuing AI systems. It includes risk management, data protection, transparency, participation, training, and incident response.
**41. Agentic AI:** AI category capable of executing sequences of actions with autonomy or semi-autonomy, triggering tools, and making intermediate decisions to achieve goals.
**42. Generative AI:** Branch of AI designed to create content, such as texts, audio, images, videos, and even code for *software* . This technology is based on advanced computational models that learn patterns from large volumes of data.
**43. Predictive AI:** System that uses historical data or observed patterns to estimate probabilities, risks, or future trends, such as dropout, performance, or need for intervention. It can support management and monitoring, but should not make automated decisions about students.
**44. Algorithmic impact:** Effect produced by an algorithmic system on individuals, groups, institutions, or educational processes, including pedagogical, evaluative, administrative, distributive, cognitive, discriminatory, or data protection consequences.
**45. Inclusion in the digital context:** Set of policies, actions, and conditions aimed at ensuring effective access and qualified use of digital technologies for all people, encompassing connectivity, devices, accessibility, digital skills, and critical appropriation. It constitutes a prerequisite for equity in the educational adoption of AI.
**46. Behavioral inference:** Automated conclusion about a student's or teacher's behavior, engagement, emotion, risk, intention, profile, or tendency based on observed data.
**47. Artificial Intelligence – AI:** technological tool dedicated to the development of computer systems capable of executing, with varying degrees of autonomy, tasks that ordinarily require human cognitive capabilities, such as perception, language comprehension, reasoning, learning, planning, and content generation, encompassing, among other applications, predictive systems, generative systems, intelligent agents, and digital tutors, with different purposes, degrees of autonomy, levels of transparency, auditability, and risk profiles.
**48. Interoperability:** Capacity of systems, platforms, and databases to operate in an integrated manner, through open standards and compatible formats, allowing information exchange, portability, and vendor replacement. It constitutes a safeguard against technological dependence and an instrument of sovereignty and educational service continuity.
**49. Digital literacy:** Set of knowledge, skills, and attitudes necessary for the critical, safe, ethical, and productive use of digital technologies, including information search and evaluation, communication, content creation, data protection, and digital citizenship.
**50. AI literacy:** Set of knowledge, skills, attitudes, and values necessary to understand, use, evaluate, contest, and create with AI in a critical, ethical, safe, and socially responsible manner. It encompasses technical, pedagogical, legal, scientific, and civic dimensions.
**51. LLM large language model:** AI model trained on a large volume of textual data, capable of processing natural language, generating text, summarizing, translating, answering questions, and supporting academic tasks. Its operation is based on probabilistic patterns and does not equate to human comprehension.
**52. Behavioral manipulation:** Use of computational techniques, including AI, aimed at hidden, deceptive, or subliminal influence on the decisions, emotions, or behaviors of natural persons, contrary to their autonomy and best interest.
**53. Data mining:** Process of exploring large volumes of data to identify patterns, correlations, and non-obvious knowledge, using statistical and machine learning techniques.
**54. Data minimization:** Principle according to which only the data strictly necessary, adequate, and proportional to the declared purpose must be collected or processed. It is an essential requirement in educational systems, especially with children and adolescents.
**55. Foundation model:** Broad model, trained with a large amount of data and capable of being adapted to various tasks, including text, image, audio, and video. It serves as the basis for multiple generative or predictive applications.
**56. Proprietary AI model:** System whose operation, weights, data, code, metrics, or documentation are controlled by a vendor, with restricted access. Its educational use requires contractual guarantees of transparency, auditability, data protection, and accountability.
**57. Content moderation:** Activity of analyzing, filtering, flagging, or removing content on digital platforms, performed by human, automated, or hybrid means. In educational environments, it must protect students against illicit or age-inappropriate content, with transparent criteria, proportional accuracy, and the possibility of human review of the measures adopted.
**58. Automated monitoring:** Use of computer systems to observe, record, classify, or issue alerts regarding the behavior, interactions, performance, or presence of students, teachers, or other members of the educational community. In the educational context, its admissibility is exceptional and subject to restrictive interpretation.
**59. Algorithmic non-discrimination:** Principle that prohibits the production, by AI systems, of unjustified unequal treatment, direct or indirect, based on origin, race, color, gender, age, disability, socioeconomic status, territory, religion, belief, or any other factor of discrimination. It imposes duties of prevention, testing, monitoring, and bias correction throughout the system's life cycle.
**60. Computational thinking:** Ability to understand, analyze, and solve problems through strategies such as decomposition, pattern recognition, abstraction, and algorithm construction, in a critical and creative way. It integrates the digital competencies provided for in the curricular frameworks.
**61. Profiling:** Automated processing of personal data to evaluate, classify, predict, or infer aspects regarding a person's performance, behavior, preferences, vulnerabilities, risks, or trajectories.
**62. Learning personalization:** Adaptation of content, pace, activities, learning paths, or feedback to the student's profile or performance. Learning personalization should not be confused with curriculum individualization or the fragmentation of the educational journey.
**63. External AI platform:** Third-party solution accessed via *the internet* , API, application, or cloud environment, without full control by the educational institution.
**64. Social scoring:** System for evaluating, classifying, or ranking people based on behavior, characteristics, or personal data, capable of conditioning access to rights, services, or opportunities. In educational systems, its use is prohibited.
**65. Natural language processing – NLP:** AI area dedicated to the understanding, interpretation, generation, and translation of human language by computer systems. It underpins applications such as *chatbots* , machine translators, spell checkers, summarizers, and large language models.
**66. *Proctoring* :**use of computer systems, with or without biometric data, to monitor students during assessments or academic activities, through the capture of image, sound, screen, environment, gaze, voice, movement, or behavioral patterns, for the purpose of identity verification or flagging suspected fraud.
**67. *Prompt* (or command):** Instruction, question, document, image, audio, or set of inputs provided to the generative AI system to request responses or tasks. It may contain context, examples, constraints, roles, and the expected format.
**68. *Prompt* hidden or *system* *prompt* :**Set of internal instructions, in rules invisible to the user, that guide the behavior, limits, tone, function, and responses of an AI model.
**69. Pseudonymization:** Processing by means of which data loses the possibility of direct association with its owner, except through the use of additional information kept separately in a controlled and secure environment. Unlike anonymization, pseudonymized data remains personal data.
**70. Tracking or traceability:** Ability to record and reconstruct decisions, inputs, outputs, versions, relevant data, parameters, responsible parties, and human interventions related to the use of AI. Indispensable for auditing, contesting, review, and accountability.
**71. Emotion recognition:** Technique that aims to infer emotional, affective, or psychological states of people from facial expressions, voice, text, physiological signals, or behavioral patterns.
**72. Facial recognition:** Biometric identification or verification technique based on facial characteristics. Because it involves the processing of sensitive personal data and captures a permanent, unalterable, and exposed attribute of a person, it is subject to a strict restriction regime in the educational environment, with exceptional admissibility and restrictive interpretation, especially when applied to children and adolescents or associated with disciplinary control, attendance tracking, or conduct management.
**73. Artificial neural network:** Computational model inspired by the functioning of biological neurons, composed of interconnected processing units organized in layers, capable of learning representations from data. It constitutes the technical basis of deep learning and large language models.
**74. Personal Data Protection Impact Assessment - DPIA:** Documentation describing personal data processing processes containing risks to civil liberties and fundamental rights, as well as the measures, safeguards, and mitigation mechanisms adopted.
**75. Human review:** Validation by a qualified person of content, recommendation, classification, or result produced by AI. It must be effective, contextualized, and proportional to the risk; it must not be merely formal, especially when it impacts rights or assessments.
**76. AI systems security:** Set of attributes that express the system's ability to operate reliably, resist failures, errors, misuse, and attacks, including command manipulation and data contamination, and to maintain adequate performance under diverse conditions. They constitute institutional adoption requirements and must be monitored throughout the lifecycle.
**77. Labeling of AI-generated content:** Clear, accessible, and verifiable identification that certain content was generated or substantially modified by an AI system, through notices, metadata, or technical marks (such as a digital watermark). It constitutes a transparency measure capable of preventing disinformation, fraud, and confusion regarding the origin of materials used in the educational environment. From the perspective of educational service provision relations, labeling is part of the basic right to adequate and clear information about the services provided, guaranteed to students and guardians as consumers of services from private institutions, and, in public institutions, stemming from the principles of publicity, transparency, and administrative good faith.
**78. Information security:** Set of technical and administrative measures intended to protect data and systems against unauthorized access, leaks, improper alterations, unavailability, and other incidents, encompassing confidentiality, integrity, and availability. It constitutes a permanent duty of educational institutions and a condition for contracting AI platforms.
**79. Artificial Intelligence System:** A computational system that, with varying degrees of autonomy and adaptability, infers from the inputs it receives how to generate outputs, predictions, content, recommendations, or decisions capable of influencing physical or virtual environments. The definition guides the identification of applications subject to the Guidelines and risk classification, distinguishing them from conventional automation.
**80. Academic recommendation mechanisms:** Mechanism that suggests content, paths, disciplines, activities, interventions, support, or management decisions based on educational data. Automated recommendation is not a neutral act: by ordering, prioritizing, or filtering options, it effectively conditions the choices of students, teachers, and managers. This is why its adoption requires a specific educational purpose, transparency about the general recommendation criteria, and preservation of the recipient's freedom of choice, prohibiting the suggestion from becoming a disguised decision through interface design, omission of alternatives, or tacit acceptance.
**81. Intelligent tutoring system:** AI-based system that offers guidance, *feedback* , monitoring, or adaptation of activities to the student.
**82. Meaningful human supervision:** Human action with effective capacity to understand, follow, interpret, contest, review, correct or interrupt the results produced by AI systems, exercised under adequate institutional conditions.
**83. Assistive technology:** Products, resources, methodologies, practices and services intended to promote the functionality, participation and autonomy of people with disabilities or reduced mobility, which may incorporate AI, such as intelligent screen readers, alternative communication and automatic subtitling.
**84. Transparency:** Duty to inform, in a clear, accessible and adequate manner for the target audience, when AI is used, its purpose, its limits, its responsible parties and its possible effects. It reaches students, families, teachers, managers and researchers, depending on the context, with language adapted to the comprehension capacity of each audience, including children and adolescents.
**85. Personal data processing:** Any operation performed with personal data, such as collection, production, reception, classification, use, access, reproduction, transmission, distribution, processing, archiving, storage, elimination, evaluation, modification, communication, transfer, dissemination, or extraction.
**86. Model training:** The process by which an AI system learns patterns from data by adjusting its internal parameters, including initial training, fine-tuning, and calibration from usage interactions. In education, the use of student, teacher, or institutional data for training requires a legal basis, a specific educational purpose, data minimization, information security, and a prohibition of exploitation incompatible with the educational relationship.
**87. Individual AI use:** Use by a student, professor, or researcher on their own initiative, outside of an institutional controlled system. Although it does not result from institutional adoption, it does not constitute a space of normative irrelevance: when carried out in a pedagogical or academic context, in the preparation of assignments, research, evaluations, productions submitted to the institution, or teaching activities, it is subject to institutional rules on the use of AI, to the duties of academic and scientific integrity, including regarding the declaration of use when required, to personal data protection rules, and the insertion of personal data of students, colleagues, or third parties, productions of others, and confidential institutional information in external platforms is prohibited, as well as the general prohibitions of the Guidelines.
**88. Institutional use of AI:** Defined, contracted, recommended, made available, or integrated use by an educational system, network, or institution. It differs from spontaneous individual use by students or teachers and attracts reinforced duties of governance, transparency, training, and responsibility. It is characterized by the institutional act of adoption, whatever its form, contracting, licensing, integration into a virtual environment, official recommendation, making access available, or incorporation into a pedagogical, evaluative, or administrative process, and the recommendation or requirement of external tool use by a teacher, in the exercise of their functions and with institutional consent or tolerance, is equivalent to institutional use for responsibility purposes.
**89. Human validation:** Procedure by which a qualified natural person invested with competence for the purpose examines and confirms, through their own reasoned judgment, the adequacy, safety, quality, pedagogical coherence, or fairness of a result produced by an AI system, as a condition for that result to produce effects.
**90. Algorithmic bias:** Tendency of an AI system, when processing data, to produce unfair, discriminatory, inaccurate, or inappropriate results for certain individuals and groups. Algorithmic bias can compromise equal treatment among students, the fairness of assessment processes, and the equitable distribution of opportunities, imposing on providers the duty to test, document, and disclose known biases and the limitations of training datasets, including regarding their suitability to the Brazilian context, and on institutions the duty to demand this demonstration during due diligence, to periodically monitor disproportionate impacts on specific groups, and to communicate relevant identified risks to the educational community, pursuant to the duty of transparency.
**91. Automation bias:** Human tendency to place excessive trust in results produced by automated systems, with a reduction in critical analysis and own judgment, leading to the acceptance of AI outputs without sufficient verification.
**92. Exclusion bias:** A qualified type of algorithmic bias resulting from the absence or under-representation of groups, territories, languages, cultures, or needs in data and models, producing inadequate, inaccurate, or discriminatory results against unrepresented populations.
**93. Emotional surveillance:** Use of AI systems to infer, classify, score, record, or monitor, on a one-off or continuous basis, the emotional, affective, or psychological states of students or education professionals, based on face, voice, behavior, writing, posture, physiological signals, or digital interaction patterns. This constitutes an aggravated form of emotion recognition, qualified by its observational and controlling nature, and subjects the intimate sphere of the person, their internal states, to automated analysis incompatible with human dignity, privacy, freedom of expression, the free development of personality, and the environment of trust inherent to the educational relationship, with the potential to inhibit the participation, experimentation, and trial-and-error that are part of the learning process.
**94. Computer vision:** Area of AI focused on the processing, analysis, and automated interpretation of images or videos, encompassing features such as object detection and classification, visual pattern recognition, document reading, and motion tracking. In the educational context, it permits applications of an inclusive and instrumental nature, such as automatic audio description, assisted reading, digitization and organization of collections and materials, but, when directed at capturing or analyzing images of people, it operates on data capable of identification and sensitive inference, approaching biometrics and surveillance, which is why its admissibility in these hypotheses is exceptional and subject to restrictive interpretation, prohibiting the conversion of legitimate purposes, accessibility, registration, or asset security into an indirect means of monitoring students and education professionals.
**ANNEX II RISK CLASSIFICATION INTERPRETATION GUIDE**
The classification applies to each specific application, functionality, and concrete purpose of use, and not to the system in the abstract or its commercial name. If there are multiple purposes or functionalities, the highest level prevails for safeguard purposes; any doubt regarding the categorization is resolved by the most burdensome classification. The expansion of functionalities, integration with institutional databases, or change of purpose mandates a mandatory reassessment. The inability to classify due to insufficient information from the supplier prevents contracting and use. In the case of children and adolescents, sensitive data, evaluation, permanence, benefits, sanctions, or certification, any doubt shall be resolved by the most protective interpretation, in light of integral protection and the best interest.
| | | | |
| --- | --- | --- | --- |
| **Risk level** | **Characterization** | **Indicative examples (non-exhaustive list)** | **Minimum safeguards** |
| I – Low | Applications for instrumental, organizational, complementary pedagogical, or accessibility support which, considering their purpose, context of use, nature of the processed data, and the system's degree of autonomy, do not produce a relevant impact on rights, assessments, academic decisions, or educational trajectories, provided they do not involve individualized profiling, processing of sensitive personal data, continuous interaction with students, monitoring, or relevant automated decision-making. | Organization, systematization, and adaptation of didactic materials; auxiliary tools for pedagogical or administrative planning; non-evaluative textual review; support for translation, summarization, and formatting; simple accessibility resources aimed at the elimination of barriers (transcription, subtitling, screen reading, magnification, format conversion), without sensitive inferences or decisional repercussion. | Transparency; information security; personal data protection; institutional accountability; simplified registration of the purpose of use; possibility of human oversight whenever used in institutional educational activities. The absence of documented classification does not authorize use under the presumption of low risk. |
| II – Moderate | Applications for pedagogical, academic, administrative, or institutional support involving continuous interaction with students, generation of academic recommendations, personalization of pathways, support for intellectual production, formative monitoring, or systematic processing of educational information, without automatic effects on grades, certification, progression, retention, benefits, sanctions, selection, or other sensitive educational decisions. This constitutes the minimum level for any application involving individualized interaction with students or systematic processing of their data. | Automated tutoring systems, prohibiting the replacement of the human teacher or tutor; formative feedback without direct academic repercussion; institutional support and guidance assistants, with clear identification of the automated nature; support for text production and research; academic recommendation without automatic or de facto decision-making effect; learning monitoring without automatic repercussion; accessibility with personalization, without sensitive data. | Clear communication to the educational community; institutional registry of applications, detailing purpose, audience, processed data, and responsible parties; age appropriateness and labeling of automated interaction; mandatory human review of results with pedagogical influence; periodic monitoring of errors, hallucinations, biases, and disproportionate impacts; prohibition of using educational data to train third-party models without an adequate legal basis; institutional channel for clarification, review, and contestation; periodic reassessment, with mandatory reclassification in the event of expanded functionalities, database integration, or purpose changes. |
| III – High | Applications capable of producing, supporting, or influencing relevant effects on students' rights, school or academic life, evaluation, certification, selection, retention, progression, retention, benefits, disciplinary analysis, or the distribution of educational opportunities, as well as those involving sensitive personal data, individualized profiling, or relevant behavioral inferences. Exceptional admissibility regime and restrictive interpretation. | Automated support for grading objective assessments with academic repercussion, exclusively as an auxiliary instrument, with individual, prior, and documented human validation; punctual identity verification and *proctoring* limited to spot checks, with a non-biometric alternative guaranteed; individualized academic profiling; systems with influence over selection, certification, benefits, retention, and disciplinary analysis, including screening and rankings that effectively condition human decision-making; processing of sensitive data; behavioral inferences and individualized prediction of dropout, performance, or conduct; detection of AI-generated content, admitted exclusively as an auxiliary element of investigation, prohibiting penalties based solely on its result; pre-correction of essay-type assessments, essays, and authorial productions, exclusively as preparatory support for the correction performed by the teacher, under the terms of art. 39, § 4. | Prior algorithmic impact assessment; data protection impact report, when applicable; demonstration of strict necessity and absence of a less invasive alternative; meaningful, continuous, qualified, and documented human oversight; traceability, documentation, and auditing; transparency and accessible explanation, including known biases; prior, individual, and documented human validation of any result with significant repercussion, prohibiting batch validation and tacit approval; contestation, full human review, and effective correction; monitoring of biases and disproportionate impacts; preservation of the non-delegable responsibility of an education professional or competent authority. Exclusively automated decisions, or those equated to it by merely formal oversight, with any significant effect, are prohibited. |
| IV – Excessive or incompatible | Applications whose potential for violating fundamental rights, discrimination, improper surveillance, exploitation of vulnerabilities, or compromise of the integrity of the educational process makes them incompatible with educational purposes and comprehensive protection. Prohibited from use under any circumstances, incapable of being validated by consent, allegedly beneficial purpose, experimental nature, or additional safeguards. | Social scoring; emotion recognition and emotional surveillance, except for the use of assistive technology exclusively for accessibility purposes, pursuant to art. 41, § 2; academic or disciplinary sanction based exclusively on the result of an AI-generated content detector; exploitation of age, emotional, socioeconomic, cognitive, or behavioral vulnerabilities, including through design aimed at maximizing engagement; psychological, biometric, or behavioral profiling for classificatory, disciplinary, commercial, or opportunity-restriction purposes; exclusively automated decisions with significant effects; continuous biometric surveillance and permanent monitoring; *proctoring* invasive, with intrusion into the private environment; use of educational data for targeted advertising, commercial exploitation, behavioral manipulation, or training of third-party models without legal basis; systems that produce segregation, stigmatization, or exclusion through automated inferences. | Prohibited use. The presence of functionality at this level contaminates the application as a whole, preventing its contracting and use, even if combined with functionalities that are permissible in themselves, and the prohibition extends to indirect arrangements that produce an equivalent result under a different name. Once the functionality is identified in an application in use, immediate discontinuation is mandatory, without prejudice to applicable liabilities. |
**FINAL PROVISIONS**
The examples contained in this Glossary and in the summary table are illustrative in nature, unless an exhaustive character is expressly indicated, and it is the responsibility of educational systems and educational institutions, in the exercise of their autonomy and governance duties, to classify the applications actually used, document this classification, and periodically review it, especially in the face of expanded functionalities, integration with institutional databases, or changes in purpose. This Glossary shall be subject to periodic updates by the National Education Council in order to keep pace with the technological, regulatory, and pedagogical evolution of the matter, ensuring the stability of the structuring concepts for the protection of rights.
**REFERENCES**
BRAZIL. Constitution of the Federative Republic of Brazil of 1988. Official Gazette of the Union: section 1, Brasília, DF, Oct 5, 1988.
BRAZIL. Decree No. 99,710, of November 21, 1990. Promulgates the Convention on the Rights of the Child. Official Gazette of the Union: Section 1, Brasília, DF, Nov. 22, 1990.
BRAZIL. Law no. 8.069, of July 13, 1990. Provides for the Statute of the Child and Adolescent and makes other provisions. Official Gazette of the Union: section 1, Brasília, DF, July 16, 1990.
BRAZIL. Law no. 8.078, of September 11, 1990. Provides for consumer protection and makes other provisions. Official Gazette of the Union: section 1, Brasília, DF, September 12, 1990.
BRAZIL. Law no. 9.394, of December 20, 1996. Establishes the guidelines and bases of national education. Official Gazette of the Union: section 1, Brasília, DF, December 23, 1996.
BRAZIL. Law no. 9.610, of February 19, 1998. Amends, updates and consolidates the legislation on copyright and makes other provisions. Official Gazette of the Union: section 1, Brasília, DF, February 20, 1998.
BRAZIL. Law no. 10.861, of April 14, 2004. Establishes the National Higher Education Evaluation System - SINAES and makes other provisions. Official Gazette of the Union: section 1, Brasília, DF, April 15, 2004.
BRAZIL. Law No. 12,965, of April 23, 2014. Establishes principles, guarantees, rights and duties for the use of the Internet in Brazil (Marco Civil da Internet). Official Gazette of the Union: section 1, Brasília, DF, April 24, 2014.
BRAZIL. Law No. 13,146, of July 6, 2015. Establishes the Brazilian Law on Inclusion of Persons with Disabilities (Statute of Persons with Disabilities). Official Gazette of the Union: section 1, Brasília, DF, July 7, 2015.
BRAZIL. Law No. 13,709, of August 14, 2018. General Law for the Protection of Personal Data (LGPD). Official Gazette of the Union: section 1, Brasília, DF, August 15, 2018.
BRAZIL. Law No. 14,533, of January 11, 2023. Establishes the National Digital Education Policy and amends Laws No. 9,394, of December 20, 1996, 9,448, of March 14, 1997, 10,260, of July 12, 2001, and 10,753, of October 30, 2003. Official Gazette of the Union: section 1, Brasília, DF, January 12, 2023.
BRAZIL. Law No. 15,211 of September 17, 2025. Provides for the protection of children and adolescents in digital environments (Digital ECA). Official Gazette of the Union: section 1, Brasília, DF, Sep. 2025.
BRAZIL. Law No. 15,388 of April 14, 2026. Approves the National Education Plan (PNE). Official Gazette of the Union: section 1, Brasília, DF, Apr. 15, 2026.
BRAZIL. Decree No. 9,235 of December 15, 2017. Provides for the exercise of the functions of regulation, supervision, and evaluation of higher education institutions and undergraduate and postgraduate higher education courses in the federal education system. Official Gazette of the Union: section 1, Brasília, DF, Dec. 18, 2017.
BRAZIL. Ministry of Science, Technology and Innovations. Ordinance MCTI No. 4,617 of April 6, 2021. Establishes the Brazilian Artificial Intelligence Strategy – EBIA and its thematic axes. Official Gazette of the Union: section 1, Brasília, DF, Apr. 8, 2021.
BRAZIL. Ministry of Science, Technology and Innovation. Brazilian Artificial Intelligence Plan – PBIA 2024-2028: AI for the good of all. Brasília, DF: MCTI, 2024.
BRAZIL. Ministry of Education. National Common Curricular Base. Brasília, DF: MEC, 2017.
BRAZIL. Ministry of Education. Artificial intelligence in Basic Education: guiding document on curricular paths and ethical practices for the use of AI in schools. Brasília, DF: MEC, 2026.
BRAZIL. Ministry of Management and Innovation in Public Services. Secretariat of Digital Government. Glossary of terms related to Artificial Intelligence. Brasília, DF: MGI/SGD, 2025.
BRAZIL. National Education Council. Basic Education Chamber. Resolution CNE/CEB No. 1, of October 4, 2022. Establishes the Norms on Computing in Basic Education – Complement to the BNCC. Official Gazette of the Union: section 1, Brasília, DF, October 5, 2022.
BRAZIL. National Education Council. Basic Education Chamber. Resolution CNE/CEB No. 2, of March 21, 2025. Establishes the National Operational Guidelines on the use of digital devices in school spaces and curricular integration of digital and media education. Official Gazette of the Union: section 1, Brasília, DF, March 24, 2025.
SCIENTIFIC COMMITTEE OF THE SCIENCE FOR CHILDREN CENTER. Study No. 15: Protection in early childhood between screens and digital media. Science for Children Center, 2025. Available at: www.ncpi.org.br Accessed on: Sep 04, 2026.
GOOGLE CLOUD. Generative AI Glossary. [S. l.]: Google Cloud Documentation, 2026. Available at: https://docs.cloud.google.com/docs/generative-ai/glossary?hl=pt-br. Accessed: 5 Jul. 2026.
MIT SLOAN TEACHING & LEARNING TECHNOLOGIES. Glossary of terms: generative AI basics. Cambridge, MA: MIT Sloan School of Management, [2025]. Available at: https://mitsloanedtech.mit.edu/ai/basics/glossary/. Accessed: July 5, 2026.
UNITED NATIONS EDUCATIONAL, SCIENTIFIC AND CULTURAL ORGANIZATION – UNESCO. Recommendation on the Ethics of Artificial Intelligence. Paris: UNESCO, 2022.
UNITED NATIONS EDUCATIONAL, SCIENTIFIC AND CULTURAL ORGANIZATION – UNESCO. Guidance for generative AI in education and research. Paris: UNESCO, 2023.
ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT – OECD. Recommendation of the Council on Artificial Intelligence: OECD/LEGAL/0449. Paris: OECD, 2019. Amended in 2024.
PORTUGAL. National Council of Ethics for Life Sciences – CNECV. Glossary on Artificial Intelligence. Lisbon: CNECV, [2025]. Available at: https://www.cnecv.pt/pt/publicacoes/estudos-tematicos/glossario-sobre-ia. Accessed: July 5, 2026.
EUROPEAN UNION. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union: L, 12 July 2024. Comparative reference.
---
| | |
| --- | --- |
| | Document electronically signed by **Israel Matos Batista** , **Advisor** , on 10/05/2026, at 18:07, according to Brasília official time, based on Ordinance No. 1.042/2015 of the Ministry of Education. |
---
| | |
| --- | --- |
| | Document electronically signed by **Antonio Cesar Russi Callegari** , **Advisor** , on 10/05/2026, at 18:26, according to Brasília official time, based on Ordinance No. 1.042/2015 of the Ministry of Education. |
---
| | |
| --- | --- |
| | Document electronically signed by **Celso Niskier** , **Advisor** , on 10/05/2026, at 18:40, according to Brasília official time, based on Ordinance No. 1.042/2015 of the Ministry of Education. |
---
| | |
| --- | --- |
| Signature QRCode | A autenticidade deste documento pode ser conferida site http://sei.mec.gov.br/sei/controlador\_externo.php?acao=documento\_conferir&id\_orgao\_acesso\_externo=0, informando o c�digo verificador **7293387**and the CRC code **5E355363** . |
---
---
| | |
| --- | --- |
| **Reference:** Process No. 23001.001062/2024-18 | SEI No. 7293387 |