AI in Schools: New York Banned It, Brazil Made Rules

In the first week of September 2026, the two largest school districts in the United States blocked students from using generative artificial intelligence. New York suspended it through 8th grade. Los Angeles blocked it for everyone, on district devices.

The day before, in Brazil, the National Education Council approved national rules for AI in education — and they follow the opposite logic.

What New York banned, exactly

The country's largest school district instituted a one-year moratorium, in effect for the 2026-2027 school year, on student-facing generative AI, from pre-K through 8th grade. It affects about 600,000 students, roughly two-thirds of the district's enrollment.

The details that matter more than the headline:

• More than 38 already-approved programs will have their AI components shut off or discontinued, for failing to meet the new safety and oversight criteria. In other words, the district didn't just block access to known tools: it reviewed what was already in the classroom.
• Companion chatbots are banned across all grades, including high school. It's the only ban with no deadline or age limit.
• In high school, rather than banning outright, the district created twice-yearly critical-thinking-about-AI modules, limited pilots in a small number of classrooms, and screen-time limits by age.

The measure was announced by Mayor Zohran Mamdani together with the schools chancellor, and the district itself describes it as the broadest in the country.

Los Angeles arrived at a similar point through a different route

At the second-largest district, the block came as a side effect of a new screen-time policy: since the start of the 2026-2027 school year, no student has access to generative AI on district devices, while the district reviews instructional use and safeguards.

It's a reversal. Previously, students aged 13 and older could use approved tools after completing a digital citizenship course and accepting the responsible-use policy.

And the district's board publicly acknowledged two limitations of the block itself: it only applies to district devices, since the district can't regulate a student's personal phone; and one board member openly said the measure was poorly communicated.

Brazil went the opposite direction: classify instead of ban

On September 1, 2026, Brazil's National Education Council approved guidelines for AI use in Brazilian schools and universities. The text now goes to the Ministry of Education for approval, and institutions will have up to 12 months to adapt after that.

The structural choice differs from the American approach: instead of banning tools, the text classifies uses into four risk levels.

• Low risk: administrative support, accessibility, text review with no evaluative purpose.
• Moderate risk: tutoring systems and formative feedback — require human oversight.
• High risk: automated grading with academic consequences and biometric-monitored exams — require continuous oversight.
• Excessive risk: social scoring, emotional surveillance, and fully automated pass/fail decisions. Not permitted.

Also prohibited: sharing identifiable student data with external systems, biometric analysis and emotional profiling of minors, and screen use in early childhood education. Rapporteur Celso Niskier's statement sums up the text's core principle: "the evaluative judgment belongs to the teacher, it cannot be replaced by AI."

What your school needs to have ready

Unlike a ban, a risk-based rule creates a documentation requirement. Under the approved text, institutions must establish:

1. An internal AI policy, stating what is allowed, by whom, and at which grade level.
2. An algorithmic impact assessment of the systems used.
3. A human oversight procedure, with a named responsible party, not just a generic mention that a teacher is in the room.
4. A review mechanism for automated decisions, for when a system fails, flags, or classifies a student.
5. Informed parental consent, specific to AI use.
6. School community participation in building these rules.

The real work isn't choosing whether to allow or block ChatGPT. It's being able to answer, in writing, which systems touch student data, what each one decides on its own, and who reviews that decision. Most schools don't even know how many AI-powered systems are already part of their routine — that's exactly what New York discovered when it found 38 programs falling short of its own criteria.

The part no rule can solve

The Los Angeles board said publicly what every principal already knows: blocking a school device doesn't stop a student from using their phone. Banning it on the network just pushes the use somewhere no one can see, with no guidance at all.

What's left is the hard part, which is pedagogical, not technical: changing what's asked of students. As long as the assignment is to produce a text a machine can generate in seconds, the debate over banning is secondary. Assessments that require process, oral defense, application to a local case, or justification of one's own choices keep working — whether AI is available or not.

That's the conversation — deciding what role technology plays inside the school — that we cover in Escola 3.0. On choosing equipment without buying what won't get used, see digital whiteboards for schools: how to choose. And on what families actually research before choosing a school, see student enrollment marketing.

Sources

New York City mayor's office announcement on the moratorium, and coverage from CNN, ABC7, and The Hill; NBC Los Angeles and GovTech coverage of the Los Angeles school district's decision; and the National Education Council's approval of AI-in-education guidelines on September 1, 2026, still pending approval by the Ministry of Education.