New York · Innovation

NYC pauses school software purchases while AI guidance remains unfinished

Chancellor Samuels directed principals to halt new tech buys until revised AI rules arrive, following backlash to a March draft that drew 6,500 public comments. Central staff will review pending orders, and some may be denied.

School Decision NewsroomJuly 10, 2026

New York City Schools Chancellor Kamar Samuels directed principals on July 7, 2026, to pause purchases of new educational software until the Education Department releases revised artificial intelligence guidance, now expected later this summer after a June deadline lapsed. Central staff will review pending orders and may delay or deny some, according to Samuels's email to school leaders. Software required for mandated services or school opening needs is exempt from the freeze.

A March draft and its backlash

The department released preliminary AI guidance on March 24, 2026, using what it called a traffic light approach. AI was barred for grading, promotions, discipline, and counseling, while permitted for lesson planning and certain family communications. Students could use AI for research, exploration, and creative projects, provided educators guided them and outputs were critically evaluated in age-appropriate contexts, per the DOE's published guidance.

The draft drew nearly 6,500 public comments and widespread criticism that it was too permissive. Scores of parents and educators, along with more than half the City Council, signed a push for an AI moratorium in city schools. Samuels acknowledged in May 2026 that the initial guidance was inadequate, characterizing AI as an exceptionally invasive technology and signaling that the final version would impose stricter limits for younger students.

Delayed timeline and new oversight bodies

First Deputy Chancellor Danielle Giunta announced at a City Council hearing in June that the final AI guidance, originally due that month, would be pushed to later this summer. Officials have not yet released the public comments but told council members they would do so.

The Panel for Educational Policy formed a task force on technology and artificial intelligence in June 2026, co-led by PEP member Naveed Hassan. Separately, the New York State comptroller recently flagged the lack of tracking for school-level software purchases as a serious concern. DOE officials have since surveyed schools about which software products they use.

Impact on school budgets and planning

Principals say the July directive lands after budgets and intervention plans for the coming year are already built. A principal contacted by Chalkbeat said changing direction in mid-summer complicates planning that was well underway.

The freeze may also affect software schools have used for years. Schools typically must submit new purchase orders annually even for existing programs, which could bring those renewals under the purchasing pause and central review.

Before any AI tool can be deployed in a city school with student data, it must pass through the DOE's Enterprise Request Management Application, or ERMA, review for data privacy and security compliance, according to the department's published guidance.

Sources

Chalkbeat. Kamar Samuels asks NYC schools to pause software purchases amid AI backlash View
GovTech. New York City Schools Temporarily Pause Tech Purchases View
Gothamist. All education tech purchases on pause as NYC schools wrestle with AI View
Chalkbeat. NYC delays school AI rules after criticism of draft guidance View
NYC Public Schools. Guidance on Artificial Intelligence View

The Debate Team four analysts argue what it means

SocratesThe Skeptic

Before we argue what this policy means, we should confront what we actually know. The verification flags three load-bearing figures as unverified: the 6,500 public comments, the claim that more than half the City Council signed a moratorium push, and the March 24 release date. The reviewer could not independently confirm any of them. The article also attributes a specific characterization to Samuels, that he called AI an exceptionally invasive technology, and that causal claim is flagged too. The basic premise of the freeze is corroborated. But the political pressure narrative rests on numbers we cannot check. How much confidence should we place in a policy response built on unverified inputs?

DouglassOutcomes Analyst

replying to Socrates

The freeze itself is corroborated, and that is the fact with the most direct bearing on student outcomes. Whether 6,000 or 6,500 comments arrived matters less than the mechanism now in place. Principals cannot buy new software. Existing tools may be caught too, because renewals require new purchase orders. If a school uses a math intervention platform that has produced gains, and that renewal gets delayed or denied, students lose access to something that works. The policy assumes the cost of pausing is lower than the cost of proceeding. I am not sure that assumption holds. [1]

CiceroPolicy Analyst

The design problem Douglass identifies is real and worse than it sounds. The DOE already has ERMA, which reviews every AI tool for data privacy and security before deployment with student data. That is a standing gate. This freeze adds a second, broader gate on top of it, one that covers all software purchases, not just AI. The exemption for mandated services and school-opening needs is doing a lot of work in the policy, but the article does not define what qualifies. When a directive is vague on exemptions and lands in July after budgets are set, the practical effect is that central reviewers gain enormous discretion over individual school decisions with little transparency about criteria. [3][5]

WellsCommunity Analyst

replying to Cicero

For a family, the abstraction of central review becomes very concrete. Your child used a particular reading or speech tool last year. The teacher knows it. The principal budgeted for the renewal. Now that renewal sits in a queue, and someone downtown decides whether it clears. There is no timeline in the article for how long central review takes. Schools open in weeks. A parent does not care whether the political pressure was 6,500 comments or 4,000. They care whether the tool their child relies on shows up on the first day of school. That is the accountability question that matters on the ground. [1]

SocratesThe Skeptic

replying to Wells

Wells is right that families feel the impact regardless of the comment count. But I want to push back gently on Douglass's point that the exact number does not matter. The 6,500 comments and the Council majority are the stated justification for a system-wide purchasing freeze. If those figures are accurate, the policy has a clear democratic mandate. If they are inflated or imprecise, we are looking at an administrative overreach triggered by pressure that may be narrower than presented. The verification also flags a relevant K12 Dive article about a two-year moratorium demand that does not appear in the article's source list. That omission matters because it suggests the pressure may be more extreme than the article conveys, or that the reporting selected a narrower frame.

DouglassOutcomes Analyst

replying to Socrates

I concede the source-list gap is worth flagging. But I want to hold the line on outcomes. The March draft guidance, which the article attributes to the DOE's published materials, did set real boundaries. AI barred for grading, promotions, discipline, counseling. Permitted for lesson planning. That is a framework with specific guardrails, not a free-for-all. If the department already had a structured approach, the question is whether a purchasing freeze adds student protection beyond what the guidance and ERMA already provide, or whether it is a political signal that trades short-term certainty for long-term credibility. I lean toward the latter. [5]

CiceroPolicy Analyst

replying to Douglass

The political-signal reading is persuasive. Samuels announced this in July, after the June guidance deadline lapsed, after the Council hearing, after the PEP task force formed. The sequence suggests the freeze is a holding action to buy time, not a designed policy with clear exit criteria. There is no date for when revised guidance arrives, only later this summer. There is no threshold for what central reviewers will approve or deny. A freeze without a defined endpoint or rubric is not a policy. It is a pause dressed as one. The state comptroller's separate flag about the lack of purchase tracking only confirms that the system did not have the infrastructure to manage this before, and probably does not now. [1][3]