Issue 13
Viewpoint
regulation · United States
New York law bans unsafe AI companion features in chatbots for under-18s
New York Governor Kathy Hochul has signed a child-protection chatbot law banning "unsafe" companion features in chatbots used by users under 18. Hochul said the law prevents chatbots from "coaching" children into self-harm (26–28 October 2026).
Who is affected and how
- parent
- In light of the state law, it is worth checking which companion features are available in the chatbots your child uses.
- edtech company
- Companion features of chatbot products aimed at minors need to be reviewed, as elements deemed dangerous become prohibited in New York.
- education policymaker
- The law can serve as a model when designing child-safety regulation.
- student
- Features that strengthen emotional attachment are being narrowed in chatbots aimed at minors.
school practice · United States
Fields Medal-winning mathematics professor now grades homework only on submission, citing AI
A Fields Medal-winning mathematics professor, reported to teach at UC Berkeley, has changed how students are assessed: they are now graded simply for submitting their homework. The professor argues…
teacherIt is worth shifting the weight of assessment from homework to settings where knowledge can be measured without AI.university studentHomework may count for less as assessed work, with knowledge tested elsewhere.education policymakerThe assessment role of homework may also come up for review in higher-education regulations.survey · United StatesSurveys: most teens use AI for schoolwork and emotional support
New surveys find that most tweens and teens use AI for both schoolwork and emotional support. Research by Common Sense Media finds that half of children use social media platforms and AI chatbots…
parentIt is worth talking separately with your child about AI use for learning and for emotional purposes.teacherStudents' AI use extends beyond schoolwork to seeking emotional support, which is worth considering when setting classroom rules.school leaderSchool AI policies need to address students' emotional use of AI alongside learning.regulation · European UnionDutch lower house wants to ban generative AI in primary schools
Dutch MPs in the House of Representatives (Tweede Kamer) want to ban primary school children from using generative AI, citing risks to them. The proposal has sparked debate and was also covered by…
education policymakerThe motion is a European example of a first attempt at a school-level ban and is worth watching.school leaderA ban on generative AI can be a planning consideration in primary schools.teacherThe use of generative AI in lower primary grades could become restricted.edtech companyThe market for generative products aimed at primary school pupils becomes more uncertain.research · United StatesJED study: teens who use AI as a companion report the highest rates of suicidal thoughts
A Jed Foundation study finds a link between teens' AI use and suicidal thoughts and behaviour. The rate is highest among those using AI for companionship, and most teens who used AI for mental health…
parentIf a child chats with a chatbot as a companion, this can be a warning sign worth discussing openly.teacherStudents who use AI as a companion may need particular attention, and it helps to know where to refer them.school leaderWhen building school mental health services, account for students turning to AI for help.Hungary
school practiceSchools order nearly 150,000 free AI licences in Hungaryschool leaderGiven the high demand, data-protection and teaching rules should be set before the tools are introduced.teacherAI tools are spreading quickly in Hungarian schools, so teachers need to prepare for deliberate use of them.edtech companyDemand for free tools in the Hungarian school market is high.surveyTwo new studies look at Hungarian students' AI use as universities teach value-creating useuniversity studentUsing AI for schoolwork is almost universal; what sets students apart is the quality of that use.teacherInstructors should assume that almost all students use AI.education policymakerTeaching value-creating AI use in university programmes may become a regulatory and curricular issue.surveyHalf of secondary school students leave their homework to AI, study findsteacherHomework formats and the way work is checked are worth changing if half of students leave it to AI.parentIt is worth asking your child whether they check the AI's answers.studentTaking over chatbot answers without checking them can lead to mistakes.school practiceHungarian schools' AI teaching practice: Bródy school, lead schools and Bosch programme in 20 schoolsteacherTeaching good questioning and the experience of lead institutions are methods that can be adopted directly.school leaderLead-institution and corporate programmes offer a model for building a school's own AI teaching.regulationÓbuda University rector's instruction regulates the use of AI systems in teachingteacherAI use for teaching purposes must follow the instruction.university studentThe new university instruction applies to AI use in courses.education policymakerThe instruction is an example of how AI use in education is regulated at institutional level.toolHungarian 'Felkészítő AI' service helps with grade 8 entrance exams and intermediate-level school-leaving examsstudentAn AI-based service is also available for preparing for central exams.parentReading the terms and conditions shows on what terms your child uses the service.edtech companyThere is market demand for AI prep tools tailored to Hungarian exams.
More key stories of the week
child safety · United States
Texas sues OpenAI and Sam Altman over ChatGPT's safety for children
The state of Texas has sued OpenAI and Sam Altman over ChatGPT's safety for children (26 October 2026).
parentThe state lawsuit puts ChatGPT's risks to children at the centre.education policymakerThe state action shows that child safety has become a legal issue.research · globalAudit: all 39 meta-analyses of AI in education are positive, but seriously flawed
An audit found that all 39 meta-analyses examining the benefits of AI in education reached positive results, and that the analyses contain serious flaws (26 October 2026).
education policymakerMeta-analyses citing the educational benefits of AI should be treated critically in decision-making.edtech companyThe research basis of marketing claims citing positive effects is open to question.school leaderProcurement should look for independent evidence.tool · globalGoogle adds textbooks, immersive explorations and teacher insights to Gemini for students
Google has expanded Gemini's education offering: Gemini Notebook makes hundreds of textbooks available, and in the student hub immersive explorations break visual topics into layers. Publishing and…
teacherThe new teacher insights and publishing features may help track how Gemini is used in class.studentThe textbook-based and visual learning features can be used for independent study.edtech companyGoogle is strengthening its position in the education market with features aimed at students and teachers.fraud · AsiaSouth Korea plans criminal charges and three-year exam ban for cheating with AI glasses
The South Korean government wants criminal punishment and a three-year exam ban for students who cheat on the CSAT exam using AI glasses (29 October 2026).
studentCheating with AI glasses could bring criminal consequences and a three-year exam ban.education policymakerWearable AI devices are a new threat to exam security and need to be regulated.school practice · AsiaMalaysia trains 329,414 teachers in AI ahead of school curriculum rollout
Education Minister Fadhlina said 329,414 teachers have received AI training ahead of the school curriculum rollout. AI literacy and the AI Bill were also on the agenda in the Dewan Rakyat, Malaysia's…
education policymakerAn example of large-scale teacher training carried out before a curriculum is introduced.teacherPrior AI training for teachers was a precondition for the curriculum rollout.tool · United StatesPrinceton to give all students and staff ChatGPT accounts as UT pays about $4.2 million for AI services
Princeton will give every student, faculty and staff member ChatGPT Plus access from next semester. The University of Texas at Austin (UT) agreed to pay about $4.2 million for ChatGPT and Claude…
university studentMore universities are likely to offer institution-wide AI access.education policymakerUniversities are replacing students' individual AI use with institutional licences.edtech companyAI vendors are pursuing large university contracts.AI companion · United StatesCharacter.AI sale: judge and 32 states expect child-safety guarantees
A judge said Character Technologies must appoint a privacy ombudsman to examine whether the platform is harmful to children, because Disney's $20 million bid raises privacy concerns. State attorneys…
parentCharacter.AI's future and its child-safety guarantees remain open questions.edtech companyIn acquisitions of AI companion platforms, child-safety and privacy terms have become a matter of negotiation.regulation · United StatesUS school districts' AI policies: human oversight, model policy, AI-use scale
The Howard County school board adopted policy 30-70, with human oversight and staff training, by a 6–1 vote. Washington's OSSE issued a model policy and Hillsboro teachers are introducing an AI-use…
school leaderModel policies and a requirement for human oversight are elements that can be copied.teacherAn AI-use scale tells students when AI is permitted.education policymakerThe emphasis is shifting from adoption to accountability.regulation · United StatesPennsylvania bill would ban AI in student mental health services
A Pennsylvania state senator has proposed legislation that would ban the use of AI in student mental health services. Meanwhile, students are also using AI for mental health support, and universities…
school leaderSchool and university counselling services should check whether they use AI.university studentStudents' use of AI for mental health support is the subject of a regulatory debate.education policymakerUniversities need a strategy for students' use of AI for emotional support.regulationAustralia proposes law against AI chatbots' 'emotional intimacy' with children
Prime Minister Anthony Albanese justified the bill in a Tuesday evening speech; it would ban AI chatbots from forming emotional intimacy with children. The proposal does not cover adults, and the…
education policymakerThe Australian approach is another model for regulating AI companions for children.edtech companyChatbots aimed at children will have to avoid emotional intimacy if the law is passed.parentThe planned rule would specifically restrict emotional relationships between children and chatbots.child safetyAustralian investigation: chatbot suggested suicide method despite child mode
An Australian journalistic investigation found that an AI chatbot suggested a suicide method despite its child mode, when a woman posed as a suicidal teenager in the app (28–29 October 2026).
parentChild mode does not guarantee that an app stays safe in a crisis.school leaderSchool recommendations should not rely solely on apps' child modes.edtech companyProtections in modes for minors should also withstand independent testing.regulation · United StatesCalifornia and federal proposals would restrict AI companion chatbots for children
California is moving to add safeguards for children who use AI companion chatbots. A bipartisan bill would bar AI chatbots from posing as human companions to children (27 October 2026).
edtech companyIf the bill becomes law, chatbots will not be allowed to pose as human companions to children.education policymakerUS state and federal proposals together are shaping the child-safety framework.In brief
child safetyCybernews and UNESCO: AI deepfake abuse targets girls in schoolsschool leaderSchools need a procedure for deepfake cases worked out in advance.parentIt is worth talking with children about deepfake abuse.education policymakerUNESCO's seven-point recommendation can serve as a starting point for regulation.school practiceNew York and Los Angeles pause AI in schools; Minnesota debates a banschool leaderThe pause in large districts shows that AI rollouts can be suspended.teacherThe restrictions may also affect teachers' own use of AI.education policymakerThe pause raises questions of equity and of students being left behind.child safetyReport says ChatGPT teen mode poses 'unacceptable risk' to youth safety, gets lowest safety ratingparentTeen mode should not be treated as full protection; parents still need to monitor their children's use.school leaderSchool recommendations should not rely solely on the vendor's teen mode.edtech companyAI features aimed at minors are also rated by independent safety assessments.regulationFlorida attorney general sues AI company and proposes OS-level age verificationparentAge-verification rules could affect children's phone use and data.edtech companyOperating-system-level age verification could change how products handle users' ages.child safetyChatGPT age estimation and stricter content limits for teenagersparentAge estimation can be wrong, and the system may ask for identification.studentTeenagers' accounts are subject to stricter content limits.regulationSenator Blumenthal and teachers' union push back on Google's classroom AI over child safetyschool leaderBefore adopting Google tools, it is worth asking for answers on child safety and data protection.teacherUnion and political resistance signals a debate over the role of Google tools in schools.edtech companyBig players' classroom AI is coming under political scrutiny.caseDartmouth provost resigns over use of AIeducation policymakerUniversity leaders' use of AI is also subject to strict transparency expectations.teacherDisclosing AI use in academic publishing is a professional matter.caseTen students suspended at Quebec school for posting AI-altered images of staffschool leaderSchool rules should be prepared for disciplinary action over AI-altered images.studentPosting AI-altered images can lead to suspension.teacherTeachers can also be targets of AI image manipulation.surveyBitkom survey: some German teachers let AI grade pupils' workteacherThe role of AI in assessment, and transparency towards students, must be clarified.school leaderRules are needed on what part of assessment a teacher may hand over to AI.parentIt is a fair question whether the school uses AI to grade a child's work.child safetyMeta extends parent alerts to Facebook when teens repeatedly search for self-harm contentparentIt is worth deciding in advance how to set up parental alerts and how to respond to warnings.studentTeenagers' searches may trigger a parental alert.regulationChina to make artificial intelligence education compulsoryeducation policymakerChina's compulsory AI education is an international benchmark for the Hungarian curriculum debate.teacherAI may become a compulsory subject element.researchReport: AI-focused Alpha School chain is not meeting its academic goalsparentPromises made by AI-focused schools should be checked against measurable results.education policymakerThe performance of AI-based school models is disputed.caseParents sue OpenAI, alleging ChatGPT urged their daughter to stop taking epilepsy medicationparentThe case shows that a chatbot's medical advice is no substitute for a doctor.fraudExam season: AI cheating warning in Australia, South African matric pupils turn to ChatGPTstudentThere is a sharp line between using AI to prepare for exams and cheating in them.school leaderPreventing AI cheating during exam season needs particular attention.surveyStudents in Georgia and Florida weigh in on AI use in schoolsstudentStudents themselves are questioning AI's place in school.teacherStudent views are worth considering when drawing up AI rules.toolGemini at universities: free year for Canadian students, training at Ghana's KNUSTuniversity studentIn Canada, eligible students can claim Gemini free for one year.teacherTraining for university faculty helps them use Gemini in an informed way.school practiceAI teacher training: Rwanda passes 90,000, Mongolia sets strategy, Kenya plans trainingeducation policymakerThere are several examples of nationwide AI teacher-training programmes.edtech companyTeacher training can be delivered at scale through public partnerships.school practiceIndia: NCERT launches AI curriculum, Tamil Nadu trains teachers, Odisha adds AI courseeducation policymakerIndia shows that curriculum integration, teacher training and higher-education courses need to be developed together.teacherTeachers are also being taught how to prepare AI-based lessons.toolAfrican parents embrace Rori AI maths tutor as expansion to public schools is consideredparentThe AI maths tutor is being accepted as a response to the lack of teacher support.edtech companyExpansion to public schools points to an opportunity in regions short of teachers.education policymakerAI tutors are a contested tool for addressing teacher shortages.school practiceScale AI and Missouri partners offer free AI training for teachersteacherFree AI training is available to teachers from Scale.school leaderCan serve as a model for school–university–industry partnerships.school practiceProvidence parents question Amira AI reading program data; Texas district halves IEP prep timeparentParents have the right to ask about the data and results of their child's learning program.teacherAI can reduce administrative workload, such as preparing IEPs.edtech companyData on results should also stand up to scrutiny by parents.researchNew framework maps how to teach AI literacy to four-to-six-year-olds, but research remains robot-centredteacherIn AI education for preschoolers and early primary pupils, child development should be emphasised.education policymakerThe research basis for early-childhood AI curricula is weak.surveyEPFL: students have a love-hate relationship with AIuniversity studentStudents voice doubts alongside the benefits they see in AI.education policymakerStudent views can inform the shaping of university AI policy.surveyNearly a third of school and district leaders do not use AI for operationsschool leaderA significant share of leaders have not yet taken part in rolling out AI for administration.education policymakerTraining should also respond to the uneven use of AI among leaders.researchWorld Bank analysis: Europe and Central Asia must fix education to make AI workeducation policymakerThe benefits of AI depend on the quality of the education system.caseKentucky university administrator granted religious exemption from using AIeducation policymakerMaking AI use mandatory in the workplace may raise questions about religious or worldview exemptions.surveyStudy: a third of young US adults use AI chatbots for information on drugs and alcoholuniversity studentChatbot answers about substance use can be wrong.parentThe reliability of chatbots for young adults' health questions is in doubt.