MEC APPROVES GUIDELINES FOR THE USE OF ARTIFICIAL INTELLIGENCE IN BRAZILIAN EDUCATION

The Ministério da Educação (MEC) ratified, on November 10, the Opinion approved by the Conselho Nacional de Educação (CNE), which establishes guidelines for the use of artificial intelligence (AI) in Brazilian education. The document sets out parameters for the adoption of the technology in education systems and public and private educational institutions, encompassing Basic Education and Higher Education.

The guidelines are intended to ensure that artificial intelligence is incorporated into teaching, learning and educational management processes in an ethical, responsible and pedagogically grounded manner. The principles include: the centrality of human governance; recognition of the value of teachers’ work; human oversight; personal data protection; information security; algorithmic transparency; inclusion; equity; the pedagogical autonomy of educational institutions; digital sustainability; environmental responsibility; among others.

One of the most important aspects of the guidelines is the adoption of a risk-based classification model for artificial intelligence applications, which establishes different governance, transparency, oversight and rights protection requirements according to the potential impacts of this technology in the educational environment.

The risk-based classification considers the purpose of the application, its degree of autonomy, the types of data processed, the potential impact on students’ rights, the level of decision automation, the possibility of profiling and the reversibility of the resulting effects. Accordingly, AI applications used in the educational environment are divided into four categories: low risk; moderate risk; high risk; and excessive risk or incompatibility with educational purposes.

Applications considered low risk include those used as tools for instrumental, organizational, supplementary pedagogical or accessibility support that do not have significant impacts on students’ rights, assessment processes or academic decisions.

Moderate-risk applications, in turn, are tools that provide pedagogical, academic or administrative support and involve continuous interaction with students, personalized learning, academic recommendations or systematic processing of educational information, without producing automated decisions that have significant effects on their academic trajectory. This category includes automated tutoring systems, formative feedback tools and institutional virtual assistants, among others. These applications require additional protective measures, such as clear communication regarding their use, institutional recordkeeping, human review of results, periodic monitoring of errors and biases, and the availability of channels for clarification or challenges to recommendations.

High-risk applications, in turn, include those capable of producing, supporting or influencing significant decisions concerning students’ rights and their educational trajectory, particularly in processes related to assessment, selection, certification, progression, retention, the granting of benefits and the application of disciplinary sanctions. In addition, AI-generated content detection tools used in academic or disciplinary investigations, as well as systems for the preliminary grading of essays and open-ended assessments, are also considered high risk. These applications require enhanced safeguards, such as prior algorithmic impact assessment, qualified and documented human oversight, transparency, traceability, institutional auditing, bias monitoring and the possibility of challenging and reviewing results.

The final category comprises applications considered excessively risky or incompatible with educational purposes. In other words, these are applications incompatible with the goals of education and the protection of students’ rights, given their potential for discrimination, undue surveillance, exploitation of vulnerabilities or impairment of educational processes. These applications are prohibited in education systems and educational institutions. Prohibited applications include: social scoring systems, emotional surveillance and automated recognition or inference of emotions; continuous biometric surveillance and permanent monitoring; psychological, biometric or behavioral profiling for discriminatory, disciplinary or commercial purposes or to restrict opportunities; among others.

The prohibitions set forth will apply immediately upon publication of the resolution, and education systems and educational institutions will have 12 months from that publication to adapt their policies, procedures, contracts and governance mechanisms. In addition, implementation must consider the particularities of each stage of education, with special attention to the protection of children and adolescents.

Artificial intelligence is increasingly present in the daily life of the Brazilian education system, transforming the ways in which educational processes are taught, learned and managed. In this respect, establishing guidelines for its use represents an important step forward in developing regulatory parameters capable of reconciling technological innovation with the protection of fundamental rights. By defining criteria for use, risk levels and oversight and governance mechanisms, the guidelines seek to ensure that AI is used in the educational environment in an ethical, safe and responsible manner, particularly with regard to the protection of children and adolescents.