Issue 5
New York City bans AI for public school students below ninth grade for one year
In early September 2026, New York City announced that about 600,000 public school students will not be able to use AI tools in the classroom below 9th grade during the 2026–27 school year. AI will be blocked or removed from school software, and screen-time limits will also be introduced; according to Politico the measure lasts one year, while Mother Jones calls it a moratorium rather than a permanent ban.
Who is affected and how
- student
- Students through eighth grade may not use AI on school devices or in school software.
- teacher
- AI-based classroom solutions must be phased out or suspended in the affected grades.
- school leader
- School leaders must review school software and enforce the new restrictions.
- parent
- Parental pressure contributed to the move, and the decision is a model example of parental advocacy.
- education policymaker
- It sets a precedent for a one-year moratorium in a large system; it is worth watching what the experience will be.
- edtech company
- Their products may be removed from school software in the affected grades.
Surveys: most US students use AI, but school rules lag behind
A survey found that between May and December 2025, US students' use of AI for homework rose from 48% to 62%, and 67% said heavier use harms critical thinking. Only about one-third said their school…
studentMost use it but suspect it harms their learning, so it is worth consciously regulating their own use.teacherTeachers cannot count on a uniform school rule, so they need to communicate their own assignment-level rules.school leaderThe absence of an institutional AI policy affects most students, so creating one is urgent.parentIt is worth asking their child about the school's AI rules.education policymakerThe gap between use and regulation justifies developing framework rules.case · globalSchool deepfake cases: Brisbane verdict, victims' accounts, prevention
A man who created sexualised deepfake images of students has been found guilty in Brisbane, and the victims described a heavy toll. Meanwhile, as the school year begins, schools and parents are…
studentYou can learn to recognise deepfakes and should know where to turn if one is made of you.teacherYou need to recognise a deepfake case and know the school's reporting route.school leaderSchool photo-use rules and procedures need to be reviewed.parentConsider which photos of your children are posted online.education policymakerDeepfake prevention should be regulated at the level of curriculum and procedure.research · globalResearch: using AI for emotional support may signal distress in young people
A University of Ottawa study found that using AI for emotional support may be a sign of distress in young people. Other research links chatbot use to greater emotional problems among teenagers and to…
studentThose seeking emotional support from a chatbot should also involve a real person who can help.teacherReliance on AI can be a sign of a student's difficulties and is worth watching for.school leaderReliance on AI should be built into the school psychology and mental-health plan.parentIt is worth talking openly with the child about what they discuss with chatbots.education policymakerSchool mental-health programmes must be prepared to recognise signs of AI use.school practice · globalUniversities move toward AI-proof assessment: oral defences, traps and rethought exams
According to Nature, instructors are responding to students' AI use with traps and AI-proof exams, and one professor caught 32 students with a hidden instruction. Denmark is requiring oral defences…
university studentCan expect oral defences and stricter exam conditions.teacherAlongside take-home essays, oral and supervised assessment is needed.school leaderExam arrangements and supervision tools need to be reconsidered.education policymakerThe Danish example shows that immediate exam-rule changes are possible at national level.edtech companyDemand is growing for proctoring and assessment tools.Hungary
More key stories of the week
Studies: AI speeds up homework, but test results get worse
Several recent reports say students who write homework with ChatGPT do worse on exams: AI speeds up homework but does not improve exam results. Bill Gates has also warned that using AI in education…
studentHomework that relies on ready-made answers does not build knowledge usable in exams.teacherHomework results alone do not show knowledge, so exam-style checks are needed.school leaderAI policies must address learning outcomes, not just cheating.parentGood homework finished quickly does not guarantee knowledge; it is worth asking for an oral explanation.education policymakerAI rollouts need to be accompanied by measurement of learning outcomes.edtech companyProduct impact will have to be demonstrated with exam-based results.regulation · United StatesLos Angeles schools also bar students from using AI on district-issued devices
The LAUSD board banned students from using AI on district-issued devices at the start of the school year. According to the US press, this means the country's two largest school districts have now…
studentStudents may not use AI on district-issued devices.teacherTeachers must align classroom AI use with district rules.school leaderEnforcing the ban and preparing future rules falls to school leaders.parentThe district will decide on AI's future role in schools with parental involvement, and parents can have a say.education policymakerThe move by two giant districts signals that moratoriums may spread.edtech companyDistrict orders and approvals may become uncertain.regulation · United StatesOpenAI backs California bill to advance youth AI safety
OpenAI voiced its support in late August 2026 for a California bill aimed at youth AI safety. Sam Altman contacted California Governor Gavin Newsom over the bill on chatbots for children.
edtech companyCalifornia's rule on chatbots for children could bring product requirements.education policymakerCalifornia's move could serve as a model for other regulators.parentThe law could strengthen protections for children using chatbots.child safety · globalChatGPT for Teens launches, but age verification faces hurdles
OpenAI introduced a version of ChatGPT tailored to teenagers in late August 2026, with curbs on ready-made answers and emotional attachment, and it is also available in Singapore. In India, the RATI…
studentAs a teenager, you may get a more restricted version built around safety.parentCheck which safeguards apply and whether they work on a shared device.teacherStudents' AI use should be assessed with the limits of the teen version in mind.education policymakerThe reliability of age verification is the key regulatory question.edtech companyThe teen version raises the safety benchmark for AI tools used in schools.child safety · globalUNICEF: 20 million young people sexually abused online in a year
UNICEF estimated in early September 2026 that about 20 million young people were sexually harassed or abused online in a year, through fake profiles, deepfake videos and photos. An estimated 1.1…
studentIf a deepfake is made of you, report it and ask for help.teacherPay particular attention to signs of online harassment and AI image manipulation.school leaderDeepfakes should be covered in the school's child protection protocol.parentTalk openly with children about fake profiles and deepfakes.education policymakerThe data argue for stricter platform and child protection regulation.tool · AsiaIndia: Bodhan AI tutor in 22 languages, IIT Madras platform piloted in Assam schools
IIT Madras's Bodhan AI has launched an AI tutor for Classes 6–12 in 22 Indian languages, and the IIT Madras platform will be piloted in 50 Assam schools after Assam signed memoranda of understanding…
studentThey can get an AI tutor in their mother tongue in Classes 6–12.teacherAlongside the AI tutor, the teacher's role is supervision and supplementing.education policymakerThe state pilot is an example of multilingual AI education rollout.edtech companyThe local-language tutor market and state partnerships present an opportunity.research · globalAI tutors not yet a replacement for humans, research says, as pilots show promise
A study finds that AI tutors do not yet replace human teachers. In an El Salvador pilot, students learning with an AI tutor reached German and Swedish levels on PISA scores, while the University of…
studentAn AI tutor can be a supplement but does not replace teacher help.teacherAn AI tutor is useful for practice, but teacher presence remains indispensable.school leaderMeasured results should be required before adoption.education policymakerThe El Salvador result may justify launching pilots, while the research argues against replacement.edtech companyThe market credibility of products without evidence of impact is weakening.school practice · globalSingapore and Australia use AI to mark essays, but teacher workload is not falling
In Singapore, a new AI tool helps mark geography essays, with English, history and social studies to follow, and Australian universities are also introducing AI-assisted marking. According to…
teacherAI marking can help, but time savings are not guaranteed and human checking remains.school leaderWhen introducing AI marking, transparency and quality control need to be planned.studentTheir work may also be assessed by AI, so it is worth knowing how.edtech companyThere is institutional demand for assessment support tools.school practice · United StatesUS districts roll out ChatGPT for teachers, including Fort Worth and St. Cloud
Fort Worth ISD, one of 55 US districts, is introducing ChatGPT for Teachers district-wide. St. Cloud, Minnesota, is also adopting a new AI tool for teachers, saying the AI will not grade students.
teacherGets an official AI tool, but grading remains a human task.school leaderA managed tool can be chosen for teachers instead of relying on individual accounts.parentThe district says AI will not grade, which can be used in communication with parents.edtech companyDistrict-wide contracts mean competition in the market for teacher AI tools.case · United StatesFederal court: AI-generated child sexual abuse material may be protected by First Amendment
A US federal court in late August 2026 dismissed the charge against a man who possessed virtual sexual images of children made with AI. The decision rests on the First Amendment and a Supreme Court…
education policymakerThe ruling points to gaps in the regulation of AI-generated child abuse material.school leaderBecause the legal background in deepfake cases is uncertain, schools need a clear procedure.parentLegal protection is not assured, so prevention and reporting matter.school practice · AsiaIsrael expands AI studies and bans phones in schools, but teachers are not ready
Israeli students have returned to school, where AI studies are being expanded and phones banned. Another analysis finds that Israel wants to lead in AI education, but its teachers are not yet ready.
teacherThere is not enough preparation for introducing AI teaching, so they need their own further training.studentThey can expect more AI studies and less phone use.education policymakerCurriculum ambitions need to be backed by teacher training.regulation · United StatesCalifornia moves to ban AI chatbot toys for children under 16
In early September 2026, California took a step toward banning AI chatbot toys for children under 16. The proposal signals a new direction in the regulation of AI products aimed at children.
parentBefore buying an AI toy, keep an eye on how the regulation develops.edtech companyThe market for chatbot toys aimed at under-16s can expect restrictions in California.