NYC Just Switched Off the AI in 38 Ed-Tech Products It Already Bought

TL;DR

On September 2, New York City announced a one-year moratorium on student-facing generative AI for every grade from 2-K through 8, covering nearly 600,000 students, two-thirds of the largest school system in the country. The same day, Los Angeles Unified confirmed that generative AI is already blocked on every district-issued device for all 378,000 of its students, a fact its own board members learned at a committee meeting. New York is not banning apps. It is disabling the AI features inside 38 ed-tech contracts it already pays for. It is also the district that lifted its ChatGPT ban in May 2023 because the original fear was, in the chancellor's words, "knee-jerk."

What New York switched off

The policy, released as official guidance on AI and screen time, draws a hard line at ninth grade. Below it, no chatbots, no AI tutors, no AI-powered instructional programs, for the whole 2026-27 year. Above it, high schoolers get access only through career and technical programs and five vetted pilots capped at 50,000 students, roughly 17 percent of the high school population. Every high schooler also sits through two 45-minute AI literacy modules a year on how the models work, bias, and misinformation.

One rule has no grade band at all: AI companion chatbots are prohibited for every student in the system. Screen time gets its own tiers. No individual devices through second grade, 30 minutes a day in grades 3 to 5, 45 minutes in middle school. Teachers keep AI for lesson planning, translation, and drafting messages home, and lose it for grading, behavior monitoring, special education plans, and anything touching student mental health.

Mayor Zohran Mamdani's line at the announcement: "Children need teachers and human connection in order to learn and grow." Chancellor Kamar Samuels was blunter: "Innovation does not mean more technology."

The kill switch is in the contract

This is the part vendors should read twice. New York is not ripping out software. Per the mayor's press conference transcript, "more than 38 previously allowed programs" will have their AI components discontinued or disabled. Chalkbeat counts those as 38 existing citywide ed-tech contracts, and named the first two: Amira, the AI reading tutor used in 222 city schools, and the AI features in Houghton Mifflin Harcourt's Into Reading curriculum, which Amira plugs into. Advocates briefed by officials say Google Gemini is being shut off on student Chromebooks. City officials would not confirm that by name, but senior advisor Ailish Brady said on the record that AI "embedded in Tier 1 curriculum where the AI can be turned off, will be turned off," and pointed at Google's products as an example.

Think of it as a building inspector who does not condemn the house but walks to the panel and pulls one breaker. The lights stay on. The one circuit the inspector does not trust goes dark until someone proves it is wired to code. The mechanism only works because those products were built with a breaker, and that is the design requirement every ed-tech AI feature now inherits: if a district cannot switch your model off without losing the rest of the product, your product is the thing that gets switched off.

Mamdani put the procurement stance in one sentence: "It doesn't matter how profitable a product is. If it doesn't help our students learn and grow, we will not use that product." Vendors that fall short "need to meet the same level of standards," he said, adding the city would work with them to get there. Enforcement, meanwhile, is delegated to principals, superintendents, and teachers, which is a polite way of saying there is no central audit of whether a product has been sufficiently scrubbed. Route Fifty flagged that gap, along with the missing full list of the 38.

Los Angeles found out at its own meeting

LA's version is broader and quieter. Until this school year, LAUSD students 13 and older who had completed a digital citizenship lesson could use approved generative AI tools. According to NBC Los Angeles and Education Week, that access is now zero on every district-issued Chromebook and tablet, for every grade, with the block enforced by the district's web filter. It came to light on September 2 at the first meeting of the board's new Generative AI Ad Hoc Committee, the body that was supposed to spend the year deciding what the policy should be. Board member Nick Melvoin's assessment: "I don't think it's well publicized." The committee meets through April 2027 and reports recommendations to the full board at the end of the school year, which means the moratorium runs at least that long.

The board's caution has a receipt. In March 2024 LAUSD launched "Ed," a custom student chatbot built by the startup AllHere under a contract worth up to $6 million, sold to the public as a "game changer." Within five months the district had pulled it, AllHere had furloughed most of its staff, and the company had collapsed, with open questions about where the student data went. A district that has already been burned by a chatbot does not need a committee to tell it to wait.

A district that has done this before

New York has run this experiment in both directions. It blocked ChatGPT on school devices and networks in January 2023, then lifted the ban on May 18, 2023, with then-chancellor David Banks writing that "the knee-jerk fear and risk overlooked the potential of generative AI to support students and teachers." Forty months later the city has landed somewhere past the original ban, because this one covers every generative tool, not one product, and because it reaches into contracts rather than a firewall rule.

The road there was not smooth. A March 2026 draft used a traffic-light system that put almost all student use in the "yellow" category and left the call to individual teachers. The backlash was immediate, Samuels conceded the draft "missed the mark," and per Chalkbeat a majority of the City Council signed a letter urging a pause and the district froze new software purchases in July. The version that shipped on September 2 is the one the critics of the draft asked for.

Not just two cities

The test for calling something a trend is more than one signal, and there are several. Fairfax County, one of the largest districts in Virginia, voted on July 16 to prohibit generative AI for all elementary students and to allow it in secondary schools only with principal or superintendent sign-off for specific projects. Newsweek's map adds Katy ISD in Texas and a Utah law that bars schools from using AI to grade student work or make high-stakes decisions. Education Week counts seven states that enacted laws on school-issued devices and digital tools in the last session, with more than a dozen considering them.

The ed-tech establishment is not thrilled. Digital Promise CEO Jean-Claude Brizard called New York's move a "knee-jerk reaction," which is the same adjective the city used in 2023 to justify going the other way. The National Parent Union's Keri Rodrigues called the rollout "a political mess" because parents "were promised a voice and got a policy handed to them." One is a complaint about pace, the other about process. Neither is a defense of the products that got switched off.

If you sell AI to schools

The five tools that made the high school pilot list tell you what the surviving product looks like. Quill is a nonprofit writing tool with a 92 percent Common Sense privacy score. Edia does math tutoring and attendance outreach for districts. Brisk Teaching, whose product is a browser extension for teachers, is on the list for teacher-created activities. Playlab and Intel's AI-Ready Schools, a curriculum program from a chipmaker, are listed for multiple subjects. None of them is a general chatbot a 12-year-old talks to alone.

• Ship the breaker. Every AI feature needs a per-district off switch that leaves the rest of the product intact. New York just demonstrated that the switch gets used, and that a product without one is the product that gets dropped.
• Teacher-facing is the market that still exists. Lesson planning, translation, differentiation, and parent communication are explicitly allowed in NYC. Student-facing tutoring is off for two-thirds of the district, and Brisk's presence on the pilot list shows a teacher-first product can still get a student-facing seat.
• The companion category is dead in schools. Not restricted by age, not piloted. Prohibited in every grade. If your roadmap had a "study buddy" persona for K-12, it now targets consumer app stores, not district contracts.
• Personal devices are the loophole both districts admit. Neither policy touches what students do on their own phones, and LA's board said so out loud. The demand does not vanish; it moves home, where the vendor is whoever the kid already has an account with.
• Expect an inventory request. Fairfax ordered a full AI inventory of its software by January. NYC found 38 contracts with AI inside. Districts are about to ask what model runs in your product, where it runs, and what it does with the transcript.

Key Takeaways

• NYC imposed a one-year moratorium on student-facing generative AI from 2-K through grade 8, covering nearly 600,000 students, and banned AI companion chatbots in every grade.
• LAUSD has blocked generative AI on every district device for all 378,000 students, a change that surfaced at the first meeting of the committee meant to write the policy.
• The mechanism is procurement, not a firewall: AI components in 38 existing NYC ed-tech contracts, including Amira in 222 schools and HMH Into Reading, are being disabled.
• High school access survives only through five named pilots capped at 50,000 students, plus two 45-minute AI literacy modules a year.
• Fairfax County, Katy ISD, and a Utah law point the same direction, and NYC itself has now gone ban, un-ban, moratorium in 44 months.
• For builders: a per-district off switch is now table stakes, teacher-facing tools are the surviving surface, and neither policy reaches the phone in the student's pocket.

Sources: NYC Mayor's Office press release, NYC Mayor's Office press conference transcript, NYC Public Schools guidance on AI and screen time, Chalkbeat New York (announcement), Chalkbeat New York (explainer), CNN, Route Fifty, NBC Los Angeles, LAist, Education Week (LAUSD), Education Week (will other districts follow), TechPolicy.Press, FFXnow, Newsweek, Chalkbeat (2023 ban), NBC News (2023 reversal), Education Week (LAUSD "Ed" chatbot), EdSurge