Meta corrects AI assistant after tool generated invasive suggestions about children

The National Education Council (CNE), linked to the Ministry of Education, approved new guidelines last Tuesday (the 1st) that regulate the use of artificial intelligence (AI) in the Brazilian educational system. One of the central rules establishes the prohibition of using AI to grade, evaluate, or assign scores for essays and discursive tests at all levels of education.

This set of norms is part of the Guidelines for the Use of Artificial Intelligence in Brazilian Education, a document that defines national standards to guide schools, universities, and both public and private education systems in incorporating this technology into learning processes.

The new guidelines also impose restrictions on direct access to AI tools for children in early childhood education and the initial years of elementary school, specifically up to the 5th grade, always requiring supervision.

The rules will come into effect after the resolution is published in the Official Gazette of the Union (DOU), granting educational systems and institutions a period of 12 months to adjust to the changes. However, the prohibitions stipulated in the opinion will take immediate effect.

Ethical-Pedagogical Objective of the Technology

According to the CNE, the purpose of these guidelines is to establish an 'ethical-pedagogical filter' for the use of technology in the school and university environment. The guidelines cover all modalities, stages, and levels of education, based on the principle that AI must serve to enhance human and educational skills, without ever replacing human judgment, pedagogical mediation, or responsibility.

Specifically, AI systems cannot be used to suggest grades or evaluate essays and discursive tests. In the case of objective tests, the technology can be used as support, but the final result must be validated by a human. Furthermore, the CNE determines that crucial decisions regarding student approval or grades cannot be made solely by automated systems.

Educational institutions must also exercise caution when using AI detectors to identify texts generated by artificial intelligence; the result of these detectors cannot be used in isolation as grounds to punish a student, aiming to prevent automatic evaluation from being the sole proof to hold a student accountable for alleged misuse of the technology.

To protect younger students, another rule reinforces that children in early childhood education and the initial years of elementary school, up to the 5th grade, cannot interact directly with AI tools without accompaniment. Institutions are also responsible for integrating teaching about AI progressively and transversally, ensuring that students acquire knowledge about the topic throughout their formation, and not just in isolation.

Additionally, schools and universities need to implement robust measures to protect student data and monitor the technology providers used in the educational environment, seeking greater control over how student information is handled by AI-based services.

Responsibilities of Universities

Universities will assume new obligations, having to create specific internal policies for managing AI in their teaching, research, extension, and management activities. This determination recognizes that technology can be applied in various areas of higher education, requiring customized rules for each context.

The guidelines also list practices that are prohibited in the educational field. In summary, the new rules define clear limits for AI in education, allowing its use only as assistance under supervision and with human responsibility. Institutions have 12 months after publication in the DOU to adjust policies and contracts, although the prohibitions have immediate application.