Should Schools Ban AI? A Better Question for Policy

On September 2, 2026, New York City announced one of the broadest student-facing AI restrictions in U.S. public education. The announcement, Mayor Mamdani and Chancellor Samuels Put Students First with Nation's Broadest Generative AI Moratorium in Schools, was issued by the New York City Mayor's Office.

For the 2026-27 school year, students from 2-K through eighth grade will not use student-facing generative AI in the city's public schools. The one-year moratorium affects nearly 600,000 students.

That sounds like a simple AI ban until you look at the rest of the policy.

High-school students will receive AI critical-thinking instruction and participate in limited pilots. Teachers can still use AI for some planning and scheduling work. The city is treating the moratorium as a period for review, not as a permanent rule for every AI use.

The details point to a more useful policy question than Should schools ban AI?

Which AI use should be allowed, restricted, tested, or prohibited, for which people, under which conditions?

Sometimes the answer to that question will still be a ban.

The technology category is too large for one rule

AI can mean many different things in education.

A student can ask a chatbot to write an essay. A teacher can use a tool to organize lesson-planning notes. A high-school class can study how a language model produces errors. An application can use AI to provide feedback. A system can generate a summary for staff.

Those uses involve different goals and risks.

A policy that says only AI is allowed leaves too much unresolved.

A policy that says only AI is banned can do the same thing unless the scope is clear.

The New York City policy is useful as a current example because it names age, student-facing use, duration, exceptions, high-school learning, and staff use separately.

You can agree or disagree with its choices. The structure of the decision is still more informative than one yes-or-no label.

Age can change the policy question

A student in early elementary school and a student preparing to graduate are doing different developmental work.

The younger student may still be building foundational writing, reading, problem-solving, and social skills. A high-school student may be ready to study how AI works, examine its limitations, and make more independent decisions about when to use it.

Age and developmental goals therefore belong in the policy.

New York City's 2026 policy makes that distinction directly. Younger students face a student-facing generative-AI moratorium. High-school students receive limited access tied to AI literacy and pilot programs.

The policy can now be evaluated against those specific goals instead of against an abstract position on AI.

Start with the learning problem

On August 20, 2026, the U.S. Department of Education released guidance on responsible education technology.

The guidance, Guidance on Responsible Use of Education Technology in the Classroom, asks five practical questions about classroom technology:

What learning problem does it solve?
When should it be used?
For whom should it be used?
For how long should it be used?
What evidence shows that it improves student learning?

Those questions apply well to AI.

They shift the burden from proving that a technology is modern or promising to showing what it contributes to a particular learning goal.

The same framework can support a restriction.

If a student-facing AI tool has no clear learning purpose for a particular age group, has unresolved privacy concerns, or replaces practice the student needs to do independently, a school can decide that the tool should not be used in that setting.

A human-centered approach makes the decision around people, evidence, and outcomes.

Privacy and safety can justify stronger boundaries

Some safety and privacy questions justify strong boundaries while learning evidence develops.

Uses can create immediate questions about data, security, age appropriateness, harmful content, or what information can be entered into a system.

A chatbot that invites a child to share personal information raises a different issue from an AI tool used by an adult to format non-sensitive notes.

A system used for grading raises different accountability questions from a system used to generate optional lesson ideas.

Specific policy makes those differences visible.

It also makes review easier. A school can ask whether the risk has changed, whether safeguards improved, and whether the original reason for the restriction still applies.

Human judgement can sit inside both adoption and restriction

British Columbia's guidance on AI in education, Digital literacy and the use of AI in education: supports for British Columbia schools, uses a human-centered approach. The guidance is published by the Province of British Columbia.

It describes potential benefits, including administrative efficiency, while stating that classroom-use choices ultimately sit with teachers based on student needs.

The U.S. Department of Education's 2026 guidance similarly emphasizes evidence, educator judgement, parent transparency, and the human-centered core of teaching.

Both approaches place evidence and professional judgement ahead of a blanket position on AI.

Professional judgement can lead to a pilot. It can lead to narrow use. It can lead to a pause. It can lead to removal when the evidence is poor.

The important part is that the decision has a stated reason and a way to be reviewed.

Assessment needs its own rule

A tool can be appropriate for one learning activity and inappropriate for the assessment of that same skill.

If the goal is to learn how to build an argument, AI might be used later to challenge the student's reasoning. During an assessment designed to show whether the student can build an argument independently, the same tool may be prohibited.

The purpose changed, so the rule can change.

Policies become clearer when they name that purpose.

Instead of asking whether AI is allowed in writing, ask what the student is supposed to demonstrate in this piece of writing and which support is consistent with that goal.

Temporary restrictions can create room for evidence

A moratorium is different from a permanent ban.

A temporary restriction can create time to evaluate tools, build training, gather evidence, define privacy requirements, and decide what age-appropriate use should look like.

A useful moratorium also needs a clear review process and deliberate learning during the pause.

New York City's policy is explicitly one year. That makes the next question important: what evidence and experience will be used to decide what happens after the moratorium?

A time limit is useful when it creates a real decision point.

Build policy around the use

A practical AI policy can ask:

Who is using the tool?
What exact task is it supporting?
What learning or professional goal does the use serve?
What human practice or judgement must remain intact?
What information can safely be entered?
How will the output be checked?
What evidence would justify expanding, changing, or ending the use?
When will the decision be reviewed?

Those questions leave room for different answers.

For some uses, the answer may be yes, with safeguards.

For others, it may be pilot it first.

For others, especially with younger students or high-risk uses, the responsible answer may be not here or not yet.

That is why Should schools ban AI? is only the beginning of the policy discussion.

The useful work starts when the technology is broken into real uses, real people, real risks, and real learning goals.

Should Schools Ban AI? A Better Question Is How It Is Used