纽约禁止六十万中小学生在校用 AI：反对的人说，这还不够

On September the second, New York City pressed pause on six hundred thousand children.

From pre-K through eighth grade, students across the city's public schools may no longer use generative AI on school grounds.

Six hundred thousand is around two-thirds of New York's public school enrolment, and New York is the largest district in the country.

In sheer coverage, this is the most thorough action any American district has taken to date.

Today we are going to take it apart: what exactly was stopped, why, and why the objections may run in a different direction than you expect.

Some background first.

Over the past two years, AI has moved into schools faster than into almost any other sector.

Student-facing products fall into roughly three categories.

One is subject tutoring — listening to a child read aloud and correcting pronunciation and word errors on the spot.

One is the companion chatbot, which a child can tell their troubles to, and which is endlessly patient and never interrupts.

The third and most widespread is the homework assistant: type in the question, out comes the answer, and it will write up your working while it is at it.

This industry has had a line for years: this is the future of education, and it cannot be stopped.

Announcing the policy, the mayor answered that line almost word for word.

He said: the tech industry wants us to believe that AI-powered early education is not only inevitable, but necessary.

He said: we do not see it that way.

The policy itself is quite specific.

The AI tools schools issued to students are taken offline wholesale, including a reading-tutor product that had been in use for a considerable time.

Companion chatbots are barred from installation on school devices entirely.

One major company's general-purpose AI assistant was switched off outright on school computers.

Alongside that, screen time was brought under management: thirty minutes a day recommended for grades three to five, forty-five for six to eight, and tighter limits on one-to-one screen time for second grade and below.

What genuinely repays thinking about is the line the policy draws: students may not, teachers may.

Teachers can still use AI for lesson preparation and administrative work, provided the tools meet safety standards.

At first glance that looks like a double standard, but the logic holds up.

A teacher's judgement is already formed, and what AI compresses for them is repetitive labour; the time saved flows back to the students.

An eight-year-old's judgement is still in the process of forming.

What they are practising at this stage is taking a thought from vague to clear, the whole way through.

Homework is only the trace that process leaves behind; the point is the process itself.

When the answer can be had in three seconds, what gets skipped is precisely the uncomfortable, and most valuable, stretch of time.

The policy leaves openings too, and leaves them with some restraint.

Students with disabilities using assistive technology are unaffected — for them, AI is an accessibility tool.

Multilingual students whose English is still at an early stage may keep using it.

Older students already taking computing courses and treating it as a career direction fall outside the ban.

Over the coming year, New York has taken on something fairly cumbersome: reviewing every technology product in use across its schools.

The threshold is whether a thing is indispensable to learning; anything below it comes off the shelf.

The review covers vendor accountability, ethical risk, instructional impact, research evidence, and real feedback from teachers and students.

Five pilot projects have been retained at high school level, running tools that have already been vetted.

Up to this point the story sounds like a familiar template: an aggressive tech industry, a cautious public sector.

But the objections come from a direction that may surprise you.

Among the critics of this policy, the loudest group thinks it does not go far enough.

A parent organisation opposed to AI in schools has said publicly that a one-year pause is too short, and that retaining high school pilots is itself dangerous.

Their reasoning: those products still have no clear evaluation of their effects, and parents lack any defined mechanism to opt out.

Seen that way, the two ends of this argument are companies saying it is an unstoppable future, and parents saying it has not been stopped thoroughly enough.

New York has placed itself in between, and given itself twelve months.

There is one more reason this is worth watching.

New York is the largest district in the country, and its purchasing decisions and classroom norms have historically been picked up as reference by other districts.

Whatever this year's review concludes is unlikely to stay inside New York.

And in education policy, a year is a short time — possibly too short to reach a conclusion, but long enough that an entire cohort will spend third and fourth grade in a classroom without AI.

That is a real controlled experiment in itself, except the subjects are actual children.

Two questions to leave you with.

The first is concrete: was there a problem in your childhood you thought about for a long time before it came clear?

If something had been sitting right there giving answers on demand, would you still have thought about it that long?

The second is bigger: when a tool genuinely does let people work faster, how do we decide at what age a child should be handed it?

There is no standard answer to these two questions, but they are worth thinking through slowly.