New York makes an AI off switch the price of staying installed in its schools

An administrator-level off switch is now a condition of remaining installed in a system that enrolls roughly 900,000 students a year ^[4], and generative AI programs already in place will be blocked or deleted from school software ^[8]. This requirement differs in kind from a usage policy. A product with AI wired through its core rather than bolted beside it does not comply by changing a setting; it complies by shipping a build that can be turned down, or it leaves. The chancellor, Kamar Samuels, issued the rule as official guidance stating that student-facing generative AI is not permitted in grades K-8 for the 2026-27 school year ^[6], and the plan needed no city council vote to take effect ^[7].

How long the posture holds is the harder question for anyone signing a term longer than a year. The school system forbade student use of ChatGPT in 2023 and reversed that ban less than a year later ^[17]. A framework published in March used a traffic-light system and permitted students to use AI for research, exploration and creative projects ^[18]; it drew anger from parents and teachers who considered it too lenient, and Samuels said leaders had "missed the mark" in their communications ^[19]. That is two changes of direction in three years ^[20]. Microsoft, which announced a partnership with New York City Public Schools in 2023 to support their exploration of generative AI tools, did not respond to a request for comment ^[24].

The scope of the prohibition is narrower than the word suggests. About 300,000 of the district's students sit outside the K-8 line ^[5]. Teachers may still use AI to prepare lessons and write messages, though they are forbidden from asking it to grade assignments ^[13], while companion chatbots offering psychological support are out entirely ^[14]. The reach of these orders also ends where a child's own device begins. Dylan Itzikowitz, a partner at South Park Commons, wrote that wealthy children will not be affected because private schools will teach them the latest technology ^[29], and the Guardian noted the unresolved problem of monitoring AI use that happens outside school ^[31].

The evidence underneath the decision is thin in both directions. Noah Fichter, chief executive of the AI startup Scout, said many school-specific chatbots being put in front of children are "actively decreasing scores" ^[27]. Sarah Shirazyan, a lecturer at Stanford Law School, said nobody knows the right approach and that the moratorium reads as taking time to understand the evidence ^[28]. Neither claim arrives with a study attached, which is roughly the case Mamdani made for pausing: officials will evaluate the policy's effectiveness after the 2026-27 school year ^[22], and he described the ban as the most comprehensive of its kind anywhere in the country ^[23].

The taskforce that will write the next version is being formed now, with teachers, elected officials, experts and union representatives ^[21]; suppliers are not one of the four groups involved ^[36]. The organized outside pressure runs one way: the AIM Coalition asked for a two-year moratorium and called this policy an important first step, adding that it is far from done with the fight ^[26]. Since New York's school technology policies are closely watched by the rest of the country ^[30], the language that taskforce settles on will travel, and the procurement conversations happening this quarter are the last ones held before it exists.