AI in schools

Brazilian Ministry of Education defines guidelines for AI in schools: what changes and how to control access

The use of AI in schools requires decisions from principals, coordinators, and IT teams: how to organize access to artificial intelligence tools while respecting the educational stage and purpose of each activity?

The Ministry of Education (MEC) approved guidelines for the use of artificial intelligence in education, with a decree published in the Official Gazette of the Union on November 11, 2026. Resolution CNE/CP No. 1/2026 covers public and private institutions, from basic education to higher education, and maintains human responsibility for decisions regarding students' educational paths.

For the school, the implementation begins with aligning pedagogical management and IT. It is necessary to define which uses are appropriate, select the services, and translate these decisions into access rules for students, teachers, and administration.

In this article, understand the points that require attention and see how to structure this control in the school routine.

What do the MEC (Brazilian Ministry of Education) guidelines establish regarding AI in schools?

The guidelines presented by the Ministry of Education address human supervision, data protection, professional training, and application governance. Technology can support educational work, but relevant decisions about students' trajectories remain the responsibility of humans.

The official statement from the Ministry of Education informs that, from the date of publication of the resolution, institutions will have 12 months to adapt their policies, practices, governance instruments, and contracts, without delaying the application of the prohibitions foreseen in the document.

The approval of the opinion and the publication of the resolution are different milestones. The counting of the deadline depends on the publication; therefore, the school should consult the final version in the Official Gazette of the Union before establishing its adaptation schedule.

CNN Brazil's report presents the guidelines approved by the Ministry of Education and the restrictions on the use of AI in the early stages of education. Watch to contextualize the topic before reviewing the differences at each stage.

YouTube player

How is the use of AI addressed at each stage of education?

The draft resolution attached to CNE/CP Opinion No. 15/2026 differentiates uses according to the stage. The table summarizes this draft; institutional application should consider the final published wording.

Teaching stages
Image 1: Teaching stages

The summary takes into account articles 21 to 25 of the draft.

For secondary education, the institution must observe the general principles of basic education and review the applicable final text, without presuming unrestricted freedom.

Can AI decide on grades, pass/fail?

The guidelines preserve human responsibility for decisions that affect students' career paths. Passing, failing, and certification should not be solely the responsibility of automated systems.

The Ministry of Education's statement clarifies that, in essays, written assessments, and other productions that require interpretation, it is forbidden to use AI as the sole tool, or without human correction, to assign grades, concepts, or merit. This does not mean a general prohibition of all technological support for assessment.

Article 40, § 4, of the draft resolution details the preparatory support: pre-correction cannot have its own evaluative effect and requires a complete reading and direct assessment of the work by the teacher. Mere confirmation of the system's suggestion does not meet these conditions.

Another restriction foreseen in the bill is the application of academic or disciplinary sanctions solely based on an AI-generated content detector. The automated result should not be treated as sufficient evidence to punish the student. This restriction is stated in article 41, item X.

For school management, this also requires reviewing the assessment resources integrated into the contracted platforms and defining how teachers will perform grading and document their decisions.

How does risk classification work for AI applications?

Articles 37 to 41 of the draft resolution organize the applications into four levels. The evaluation depends on the purpose, the data used, and the effects on the student.

Risk level
Image 2: Risk level

This classification is not a list of permitted or prohibited websites. The same platform may combine functions with different risks; the institution needs to assess the intended use before defining access permissions.

Do schools need to block ChatGPT and other AI tools?

Approval, in itself, does not allow us to conclude that all schools need to block all AI services for all audiences. The technical decision must reflect the educational stage, the tool, the activity, and the applicable restrictions.

A teacher preparing materials and a student using a chatbot on a shared computer are different situations. Applying a single internet policy to both could allow inappropriate access or hinder authorized activities.

It is also necessary to differentiate between the types of resources. A chatbot that produces text, an educational recommendation system, and assistive technology have distinct functions and risks. The presence of AI in a service requires an evaluation of its specific use.

Blocking access can help implement a pedagogical decision, but defining that decision rests with the institution.

Before releasing a tool, it's worth answering:

  • Who will be able to use it and at what stage of education?
  • What educational objective justifies access?
  • How will mediation take place, when necessary?
  • What data will be processed?
  • Who approves, monitors, and reviews its use?

How to transform guidelines into AI access policies

An applicable policy needs to link the pedagogical rule to the environment in which access occurs. This requires combining people, devices, networks, and responsibilities.

1. Map the services and access environments

Start with the resources already used by teachers, students, and administration. Include purchased tools, free accounts, and AI features integrated into educational platforms.

For each service, record the purpose, the target audience, the data involved, and the internal person responsible. Then, identify where access occurs: laboratory, institutional devices, faculty network, or personal equipment.

This survey prevents the school from concentrating control on a well-known chatbot and neglecting to evaluate other services used daily. It also helps identify tools used without institutional approval, one of the risks of Shadow AI and data leaks, discussed on the Lumiun blog in the context of businesses.

2. Separate the profiles that need different rules

The organization can begin with policies for students, teachers, administration, and visitors. When necessary, students can be divided by grade level or by area of ​​use.

This separation needs to exist technically. If teachers and students use the same network without identification or differentiation of devices, a single policy applied to the connection will have difficulty representing their needs.

The IT team must choose the method compatible with the infrastructure: separate networks, device groups, or user identification, depending on the available resources. The article on access control by groups and devices with ActiveNet explains how this organization allows applying different policies to each profile.

3. Define how release requests work

The school needs a clear path for evaluating tools. A teacher's request can specify the class, activity, service, duration, and necessary data.

The coordination team evaluates the pedagogical proposal. The IT department verifies the access requirements and configuration. The privacy analysis team should be involved when personal data is processed.

Documenting this workflow reduces improvised approvals and helps to review permissions that have lost their purpose.

4. Organize the use of shared laboratories

Imagine a laboratory used in the morning by a 4th-grade class and in the afternoon by an 8th-grade class.

If an AI activity is authorized for the second class, permanently releasing the service to the entire lab could also make it available to other students.

One option is to prepare a group of devices or a specific environment for the activity and apply the corresponding permission. At the end, the responsible team verifies whether access should be terminated.

This is a suggestion for technical organization. The release of the class remains dependent on the pedagogical assessment and the rules applicable to the class.

5. Test the policy before the activity

Verify the allowed and blocked accesses on a device with the same profile as the students. Some tools depend on other domains for authentication and operation.

The test should answer two questions: Does the authorized activity work? Did the exception open up other accesses that the school intended to restrict?

Also define who can help during the class and how to request adjustments.

How Lumiun DNS can help control AI tools

Lumiun Lumiun DNS is a DNS security and browsing filtering solution that allows you to apply internet access policies.

Among the features is the Artificial Intelligence, aimed at blocking generative AI services. The platform also allows you to unblock selected services through blocklists and permission lists, according to the configuration described in the Help Center.

Lumiun DNS Panel - AI
Image 3: Lumiun DNS Panel – Artificial Intelligence Blocking

With distinct policies for each environment, group, or device, depending on the adopted configuration, the school can translate its decisions into navigation controls. Some examples:

Environment Possible application
Student devices Restrict generative AI services according to the policy defined for that audience.
Authorized activity environment Allow evaluated services for the group or devices involved.
Teachers Apply permissions compatible with the approved tools for planning and pedagogical work.
Administration Maintain your own rules for services used in administrative activities.

DNS query and blocking reports help review policies and identify settings that need adjustment. However, a query to a domain does not prove that the student interacted with a chatbot or submitted information.

What are the limits of DNS filtering?

The DNS filter acts on the domains requested by the devices, applying blocking or allowing rules during name resolution. It does not identify the presence of the teacher, does not analyze the content of each conversation, and does not determine if a generated response is appropriate for the activity.

If a platform brings together different functions in the same domain, it may not be possible to separate only the AI ​​function through this control. Similarly, devices using external connections or other paths outside the managed configuration require further evaluation.

Therefore, the filter must be integrated into the institution's set of controls, along with supervision, device configuration, and tool evaluation.

What else does the school need to organize besides the lockdown?

Even a tool permitted on the network still needs rules for its use. As a management practice, the school can establish guidelines on data, content review, authorship, and communication with families.

Protection of data used in activities

Before adopting a service, evaluate what information it requests, how it stores it, with whom it shares it, and whether it uses it for model training.

In activities, prefer fictional examples when real student data is not necessary. Instruct professionals to consult the data protection officer before entering identifiable information into external tools.

These precautions are in addition to student guidance and device protection, discussed in the online safety guide for educators.

Teacher preparation

The Ministry of Education's statement foresees ongoing training for education professionals. In practice, this training can address concrete situations: verifying an incorrect answer, recognizing a non-existent reference, discussing biases, and reviewing material produced with the aid of AI.

It is also helpful to explain how to request a tool and what conditions need to be met to use it with a class.

Communication and review of the rules

Students and families need to understand what the institution allows, in which activities, and for what reasons. Guidance should be age-appropriate and consistent with classroom practice.

This concern is part of a broader discussion about protection in the digital environment, addressed in the article "Digital ECA: what changes for schools and IT".

The rules should be reviewed when a service changes its function, begins to require new data, or ceases to meet its approved objective.

Where should I start at your school?

Bring together the pedagogical coordination and IT teams to identify the AI ​​services in use and choose an initial environment for implementing the policies. Define the target audiences, those responsible for approvals, and the review criteria.

Next, test the configuration with the professionals involved and record any necessary adjustments before expanding it. This sequence helps maintain consistency between the school's guidelines and the access available in practice.

Does your institution need to implement different browsing policies for students and teachers? Learn about Lumiun DNS for educational institutions and see how to organize access control on the school network.

Try Lumiun DNS for free
Related Posts