AINEWS
New York's classroom AI moratorium stops at the school door. Korea's screens have lived on the dinner table all along, so copying the rule leaves the problem exactly where the child is.
The Korean restaurant on a Saturday evening: a smartphone propped in front of a 5-year-old so the parents can pick up their chopsticks. The phone was not handed over because the child begged, nor out of neglect. It is the tool an adult used to buy thirty minutes of quiet.
That scene is the gap that New York's classroom AI ban, announced on Sept. 2, 2026, cannot reach. The moratorium drew a clean line at the schoolhouse door. In Korea, the screen is not at the schoolhouse door. It has been on the dinner table, the back seat, and the stroller all along.
The New York City Department of Education's Sept. 2, 2026 moratorium covers roughly 600,000 public school students from 2-K (2-year-olds through kindergarten) through Grade 8, about two-thirds of total enrollment, for one year. Generative AI is off-limits in that span. Companion chatbots, the kind that chat with a child, are banned across all grades. Five vetted high-school tools get a tightly capped pilot: up to 50,000 students, five classrooms per school, ten to twenty minutes per tool per week. The pilot list is Quill for close reading, Edia for math, Brisk Teaching across subjects, Playlab for dissecting AI bias, and Intel AI-Ready Schools as a semester project program.
The carve-outs matter. Teachers can use AI for admin, lesson prep, and translation. They cannot use it to grade work or counsel students. The Department layered screen-time caps on top: one-to-one devices banned for Grade 2 and below, thirty minutes a day for Grades 3-5, forty-five minutes a day for Grades 6-8. Official AI education is a forty-five-minute critical-thinking module, twice a year.
Korean first graders finish around 1:00 p.m. Parents work much later. The gap between dismissal and pickup is the gap the state built Neulbom School to fill: expanded public after-school care designed for working families. Before Neulbom, elementary after-school program participation sat at 50.3% and elementary care classes at 11.5%. The phone on the restaurant table is the same gap, at smaller scale. It is childcare labor outsourced to a device.
The 2025 Smartphone Overdependence Survey from the Ministry of Science and ICT, released March 2026, puts numbers on the household side. Overall, 22.7% of Koreans fall in the overdependence risk group, a fifth straight annual decline. The cohort that worries the ministry is the youngest: 26% of children aged 3-9 are in the risk group, up from 17.9% in 2016. For adolescents 10-19 the figure is 43%.
A classroom rule does not touch those numbers. The screen is in the household because that is where the time is.
Korea tried the classroom version of this experiment and watched it collapse. The AI Digital Textbook program launched in March 2025 with a ₩1.4093 trillion (about $1.01 billion) three-year budget, covering 3rd-4th grade elementary and 1st-year middle and high school for English, math, and informatics. By the second semester of 2025, participating schools had dropped 58.8%, from 4,095 to 1,686. An Aug. 4, 2025 amendment to the Elementary and Secondary Education Act downgraded the AI textbook from official textbook to instructional material. In 2026 it became a discretionary "AI learning aid," and student utilization, disclosed in a parliamentary audit, sits at 8.1%.
When the public system turned it off, private spending turned back on. The 2025 primary and secondary private education spending survey came in at ₩27.5 trillion (about $19.8 billion), the first decline in five years, but average monthly spending per participating student rose 2.0% to ₩604,000 (about $432), crossing the ₩600,000 threshold for the first time. Total participating students fell 2.3% to 5.02 million, fewer students, more spent on each.
Conversational AI is two-way. It does not scold. It plays along with child-made rules, so parents read it as more controllable than a YouTube autoplay queue. A prior INLEVEL9 issue found a 13-percentage-point gap between teen chatbot usage (64%) and parental awareness of that usage (51%), with 16% of teens using chatbots for everyday conversation and 12% for emotional support. The phone on the table is one screen. The chatbot is the one the parent cannot see.
Korea's regulatory sequence has run the other way around. In-class smartphone possession was banned starting in March 2026. The Korea Media Commission briefed a policy restricting social sign-ups for under-14s and algorithmic recommendations for 14-19s. The June 2026 local elections surfaced mandatory parental consent for under-16 social users as a top pledge across metropolitan and provincial superintendent candidates. The Ministry of Education is piloting "smartphone-free schools" in the second semester of 2026. Each step is a classroom-shaped rule aimed at a household-shaped problem.
Australia is the cautionary case from the other direction. The teen social media ban reportedly led to about 4.7 million under-16 accounts being deleted, and a parent survey found 69.1% still reporting that their child had an Instagram account. Forty randomized trials have tested whether quitting social media improves mental health. None had an average sample age under 18.
The strongest objection to this analysis is the partial one: a classroom ban still reduces in-school exposure. That is real. It addresses a fraction of the day. It addresses none of the reason the screen is in the household in the first place.
The screen, in Korea, was a childcare tool from the start. New York's moratorium addresses the moment. Korea's problem lives in the hour after the bell.