Issue 4
Viewpoint
child safety · global
OpenAI launches ChatGPT for Teens with study features, parental controls, for ages 13–17
OpenAI on Tuesday, 25 August launched a version of ChatGPT for 13–17-year-olds, with stronger content filtering, parental oversight and study-support features. When a teen asks for a quick answer, the system steers them toward thinking it through; it arrives in Australia in September.
Who is affected and how
- student
- Students use ChatGPT in a separate, filtered environment that can ask guiding questions instead of giving a ready-made answer.
- parent
- Parents should review the parental-control settings and discuss them with their child before the child starts using it.
- education policymaker
- Age-appropriate chatbots set a new benchmark for shaping school AI rules and procurement.
- edtech company
- Default safety protections and learning-focused design are likely to become the expected standard for products aimed at teenagers.
tool · United States
OpenAI extends ChatGPT for Teachers to 55 more US school districts
OpenAI on 26 August extended its programme for teachers to a further 55 districts, giving more than 100,000 additional educators access. The programme is free for verified US K–12 educators; partners…
teacherIn the districts concerned, verified educators can use the tool free of charge for planning and creating materials.school leaderDistrict partnerships need to be accompanied by rules on use and data handling.education policymakerDistrict-level agreements offer a model for introducing an AI tool at system level.edtech companyOpenAI is embedding itself in teachers' workflows through district partnerships, which sets a benchmark for competitors too.research · United StatesStudents rarely use AI tutors, even with human help, without motivation, studies find
Chalkbeat reports that students rarely engaged with Khanmigo, and several analyses conclude that willingness to use AI tutors limits their impact. The tools can help students learn new material, but…
teacherThe AI tutor has to be built into the lesson and its use encouraged, otherwise students will not use it.school leaderBuying licences does not deliver results by itself; actual use and motivational factors also need to be measured.edtech companyIn product design, involving students and motivation matter at least as much as content quality.education policymakerProgramme evaluation should measure actual use rather than access.research · United StatesStanford review finds AI tutoring works best when human tutors are involved
A Stanford review concludes that AI-based tutoring is most effective when human tutors are involved, and does not replace the tutor. The research base for high-impact tutoring is already strong, and…
education policymakerIn tutoring programs, AI should be funded as a complement to human tutors, not as a replacement.school leaderRollout should provide human supervision and contact alongside the AI tool.edtech companyThe evidence favors hybrid products that support the human tutor.research · United KingdomStudies: generative AI cannot grade essays accurately and tends to give higher marks
Cardiff University says generative AI cannot accurately grade higher-education essays, and another study found that AI tends to give higher marks than humans do. According to the reports, teachers…
teacherAI grades should not be accepted without checking, because they are inaccurate and can skew upward.university studentAI marking may be more favorable, but it is not necessarily accurate feedback.education policymakerBefore AI grading is introduced, rules for human review and calibration must be set.edtech companyAutomated grading products must demonstrate accuracy and freedom from bias.Hungary
surveyThree-quarters of Hungarian families already use AI for learning, MediaMarkt–Europion survey findsparentAI use for learning is widespread, so parental supervision and shared rules should be actively developed.teacherAI use at home is widespread, so it must be factored into planning homework and assessment.education policymakerRepresentative Hungarian data is available to underpin regulatory and school guidance.studentAI is a normal tool when studying, while parents increasingly want to supervise its use.toolGoogle offers Hungarian university students a free year of paid Gemini planuniversity studentWith proof of student status, the paid Gemini plan is free for one year; it is worth weighing the costs that follow after the year ends.education policymakerSome university students will work with Google's tool, and institutional AI rules need to cover this as well.edtech companyFree large-model bundles put local providers under strong price pressure in the student market.toolHungary's BOOKR Kids closes €6.1 million Series A for international expansionedtech companyA Hungarian edtech company has raised significant funding for international scaling, setting a market benchmark for local players.parentAI-based reading and learning products for children are spreading; check the data handling and content before use.regulationNew leadership post to be created at Hungary's Educational Authority, with AI among its taskseducation policymakerAI use and data analysis become formal tasks at the Educational Authority, which may also signal the future policy direction.school leaderData-driven and AI-related expectations are likely to start coming from the authority as well.researchEducatio study examines regulation of AI use in Hungarian higher educationeducation policymakerA Hungarian reference point for comparing and developing institutional AI regulation.university studentThe study may show how much universities' rules differ, so it is worth checking each course's requirements separately.caseTirana installs AI cameras in schools under the Smart City programmeparentParents should find out how data from school AI cameras is handled.school leaderIntroducing AI-based surveillance raises data protection and ethical questions.studentStudents' movements and behaviour may be tracked by AI cameras.school practiceProfessional programme on conscious AI use launched for teachersteacherA professional training opportunity for using AI consciously in lessons and planning.school leaderThe programme may be one available way to prepare colleagues.toolUniversity of Szeged and Zengo Kft. develop AI-supported VR/AR education platformedtech companyAn example of university–company cooperation on Hungarian AI-based educational development.teacherA new VR/AR-based way of creating teaching materials may emerge in education.
More key stories of the week
case · United States
Teachers become targets of students' AI-generated sexual deepfakes
According to accounts from US and UK teachers, students are using AI to create sexualised deepfake images of them, testing the Take It Down Act. Schools have no clear playbook, and WIRED presents the…
teacherIn cases of deepfake harassment, document and report it, and ask the school for protection.school leaderWithout a procedure prepared in advance, the school is exposed; this gap must be filled.education policymakerRegulation does not cover deepfake harassment committed by students.studentCreating deepfakes can have serious consequences.child safety · United StatesCritics demand proof and further regulation of ChatGPT for Teens safety
After the launch, critics, including a mother who lost her son to a chatbot, say OpenAI must prove the safety it claims. TechRepublic points to gaps in age estimation and privacy, while opinion…
parentParental controls do not replace conversation; given the gaps in age estimation and privacy, a child's use should still be monitored.education policymakerGiven the lack of independent evidence and the identified gaps, regulatory requirements (a safety standard, verifiability) should be spelled out.edtech companySafety claims for AI products aimed at children must be backed up in a measurable, verifiable way.regulation · United StatesUS appeals court rules AI-generated child abuse images depicting no real child are protected at home
A federal appeals court, the Seventh Circuit, has held that the First Amendment protects possessing AI-generated child sexual abuse images at home when they depict no real child. A federal judge…
education policymakerBecause of this legal gap, school and child-protection rules cannot rely on criminal law alone.parentDespite the ruling, deepfakes of real children remain a serious risk; it is worth discussing with the child.school leaderSchools need an internal protocol for handling AI-generated explicit content.tool · United StatesKhan Academy adds visual tools and teacher controls to its AI tutor
Khan Academy updated its AI tutor with visual tools and teacher controls, according to reports on 27 August. The update is meant to strengthen the teacher's role in the system.
teacherThe new controls give teachers more say over what the AI tutor does in class and how.school leaderThe new features favour introducing an AI tutor that operates under teacher oversight.school practice · AsiaSingapore universities review AI detection policies as SUSS drops its detector
Singapore universities are reviewing their use of AI-detection software because they doubt its reliability. SUSS has dropped its AI detector.
university studentA detector flag alone cannot be evidence, but it is worth documenting the process behind your work.teacherInstead of disciplinary proceedings that rely on a detector, process-based assessment is advisable.education policymakerThe role of detectors needs to be rethought in institutional policies.fraud · AsiaSouth Korean exam cheats switch to AI-powered smart glasses
According to a report of 25 August, exam cheats in South Korea are using smart glasses with AI capabilities. The technology poses a new challenge for exam supervision.
education policymakerExam security rules must also cover wearable AI devices.teacherIn supervision, smart glasses must be checked alongside traditional phones.studentUsing such devices counts as exam cheating, with consequences.survey · globalOpenAI says ChatGPT sees 70 million learning conversations a week
An OpenAI report says ChatGPT hosts 70 million learning-related conversations a week, involving students and teachers. The figure was published on 26 August and picked up by the Hungarian press.
education policymakerThe vendor's own figure shows scale but says nothing about learning outcomes; independent measurement is needed alongside it.teacherStudents are already using the tool for learning in practice, so lesson planning must take this into account.regulation · United StatesUniversity of Chicago bans technology and AI in social sciences core courses
The University of Chicago's social sciences division is switching to an "analog" mode this fall and banning technology and AI in core courses. According to the Chicago Maroon, the ban applies…
university studentIn the affected core courses, students may not use AI or digital devices in class.teacherCourse policies and assessment methods must be adapted to the ban.education policymakerThis is an example of different AI rules coexisting within one institution.regulation · United StatesFlorida considers AI rules for state colleges
According to news from 26 August, Florida is planning rules on AI use at state colleges. Details are not yet included in the reports.
education policymakerState-level AI rules for colleges are being prepared and could serve as a model.university studentThe rules may also affect the conditions for using AI at colleges.child safety · United StatesxAI accused of training Grok's deepfake engine on real child abuse material
According to IBTimes, Elon Musk's xAI is accused of using real child abuse material to train the deepfake engine of Grok. A New York law firm has published an explainer on the lawsuit for the parents…
parentParents of victims may also consider legal action; understanding the lawsuit helps in deciding.education policymakerThe provenance of training data is a child-protection regulatory issue.edtech companyTransparency about training-data provenance and the handling of generated images is an increasing risk.AI companion · United StatesAI chatbots as therapists and companions: who new state laws protect and what experts tell parents
New state laws govern whether AI chatbots can be used as therapists or companions, and whom they protect. Pennsylvania experts are warning parents that children are relying on chatbots for therapy.
parentIt is worth discussing a child's emotional conversations with a chatbot and steering them towards a professional.education policymakerState laws provide a framework, but school rules must also address protecting children.edtech companyTherapy-like or companion features carry state regulatory risk.toolArmenia to provide 50,000 ChatGPT Edu subscriptions to schools and universities
Armenia is providing 50,000 ChatGPT Edu subscriptions to schools and universities. From September, educational AI tools will be available in Armenian schools and universities.
education policymakerA nationwide AI licence is an example whose lessons may prove instructive.teacherFrom September, AI tools will be available in education.studentAccess in schools and universities may be nationwide.In brief
caseYale lawsuit shows an AI-detector flag alone does not hold upteacherA detector result is not enough for an accusation; further evidence and a hearing must be provided.university studentIf AI use is wrongly suspected, documenting edit history and workflow can help.school leaderDisciplinary rules must be made legally defensible given the limits of detectors.researchMIT committee: AI forces a rethink of nearly every element of educationteacherIt is worth rethinking assessment formats (non-traditional grading) and social learning.education policymakerA detailed AI strategy report from a leading university can serve as a reference.university studentAssessment and learning formats may change.toolUniversity of Colorado Boulder and California State University give students ChatGPT accessuniversity studentWith institutional access the tool is free to use, provided students know the rules and the data handling.education policymakerThe results and concerns of large-scale licences are worth examining before a decision.edtech companyCampus-wide university licence agreements give a competitive advantage.toolAnthropic launches Claude for Teachers for US K–12 schoolsteacherAnother AI assistant option has appeared for planning and creating materials.school leaderBefore procurement, compare the offerings on data protection and regulation.edtech companyMajor model developers are entering the teacher market directly, which puts pressure on standalone edtech tools too.surveyTeachers' AI use soars but has little effect on working hours, Tes reportsteacherAI tools do not guarantee less work, so they are best used in a targeted way.school leaderReducing teacher workload should not be expected from AI adoption alone.education policymakerManaging teacher workload requires further measures.surveyCanadian survey: four in 10 teachers in grades 6–12 say over half of students use AI for homeworkteacherIn-class, supervised written tasks may play a bigger role as homework becomes less reliable.education policymakerTeachers' experiences justify rethinking assessment and curriculum directions.parentHomework may be done with AI help, so it is worth clarifying expectations with the child.regulationMeta's $17 billion child-safety settlement is the biggest tech payout ever, Fortune reportseducation policymakerThe settlement shows the heavy financial consequences of child-safety failures.edtech companyThe legal and financial risk of safety failures in products aimed at children is significant.regulationIdaho makes AI policies mandatory for schoolsschool leaderSchools and districts must draw up their own AI policies.education policymakerThe state-level mandate is an example for other states.caseNew York Times: how Big Tech captured American schoolsteacherCommercial interests also lie behind training sessions organised by companies.school leaderProcurement decisions should weigh independent evaluation and the risk of dependence.education policymakerThe influence of tech companies in schools needs to be handled transparently.school practiceMost Israeli students already use AI; schools aim to teach how to use itteacherThe curriculum needs to teach when to use AI and how to question it critically.education policymakerA state-level AI education programme offers an example, with big tech companies involved.studentDeliberate judgment about AI use becomes part of learning.regulationUS school districts draft AI policies; Norwalk wins $23,000 grantschool leaderSupport and workable processes are available for drafting policies.education policymakerExternal organisations also provide funding for developing district rules.school practiceHong Kong holds AI leadership training while DepEd and Microsoft expand AI in Philippine schoolsschool leaderIn Hong Kong, AI leadership training is a tool for developing digital education in schools.education policymakerThe partnership with Microsoft offers a model for large-scale AI rollout.school practiceSchools in Canada and the US teach AI literacy: mandatory course in Winnipeg, critical thinking in LincolnteacherTeachers need to prepare to teach AI literacy, encouraging critical use instead of banning AI.school leaderExamples of mandatory courses and curriculum integration are available.studentAI literacy is becoming formal course content.AI companionOpinion: banning AI toys for four years will not make children saferparentWhen buying an AI toy, parents cannot count on a ban and must judge the product's safety themselves.education policymakerTargeted safety requirements could be considered instead of a ban.edtech companyThe regulatory direction for AI toys is uncertain.researchBocconi study with OpenAI finds ChatGPT use and critical-thinking training have complementary effectsteacherCritical-thinking training should be built alongside the use of AI.education policymakerThe research was conducted jointly with the vendor, so independent verification is also needed.school practiceCornell launches AI critical literacy program for all first-year undergraduatesuniversity studentFirst-year students will have to take a program on the critical use of AI.education policymakerAn example of a university-wide AI literacy model that covers every student.school practiceCommon Sense Media completes K-12 AI literacy curriculumteacherA ready-made curriculum is available for teaching AI literacy.school leaderA ready-made framework can be chosen for introducing AI literacy.surveyOver 70% of students use AI for study every week, Cork Independent reportsteacherTeaching and assessment need to be adjusted to AI use as a common practice.education policymakerThe high share makes the development of guidelines more urgent.school practiceAI prompts Dutch secondary schools to rethink homeworkteacherThe form and assessment of homework need to be redesigned in light of AI use.school leaderSchool-wide homework rules may become necessary.surveyReport: only 13% of Indian schools are AI-readyeducation policymakerThere are serious gaps in infrastructure and readiness, so targeted investment is needed.school leaderSchools' readiness should be assessed before AI is introduced.caseMan accused of making deepfakes of students said it was 'like X-ray vision'parentFake explicit content can be created from children's photos, so photo sharing is worth reviewing.school leaderSchools need precautionary rules for publishing school photos.caseTeacher says IDEA Public Schools fired him over 'deepfake AI' claimteacherA clear procedure and examination of the evidence can protect against wrongful dismissal in deepfake claims.school leaderDeepfake claims must be handled in a verified process before any employment action is taken.school practiceChina builds AI-driven education, from textbooks to robotseducation policymakerAn example of a national, AI-centred curriculum model.teacherChinese practice reflects an international trend toward teaching AI.researchBrookings: AI surveillance in schools raises safety and equity concernsschool leaderBefore buying an AI surveillance system, ask for evidence of its effectiveness.education policymakerEvidence-based procurement conditions should be developed.studentAI surveillance may disproportionately affect certain groups of students.caseAt a Norwegian university, examiner who suspected a student of AI cheating broke the AI rules themselvesteacherAI rules must be enforced on assessors as well.university studentIf suspicion is mistaken, the assessor's rule violation may be grounds for appeal.research6,997-student trial: AI tutor slows math practice but aids recovery after mistakesteacherThe tool can help students learn from mistakes, but the pace of practice needs to be monitored.edtech companyThe dual effect measured in a large sample is a reason to measure the effects on speed and on learning separately.regulationAI ethics group urges rethink of AI grading for South Korea's college entrance exameducation policymakerFor a high-stakes exam, the transparency and control of AI grading must be clarified.studentAI grading of exams raises questions of fairness.toolSaxony-Anhalt introduces AI chatbot AIS.chat in schoolseducation policymakerAn example of a state-level, government AI chatbot rollout.teacherA new official AI tool may become available in schools.regulationEU moves to tighten children's online protection, with addictive design in the spotlightparentEU regulation may strengthen children's online protection.education policymakerEU directions will affect national child protection rules.edtech companyAvoiding addictive design may become a regulatory expectation.caseDutch Supreme Court rejects appeal in case over student's ChatGPT useuniversity studentDisputes over AI use can also be settled through legal channels, and rulings may set precedent.education policymakerThe final ruling offers guidance for handling AI cheating cases.survey35-country survey asks who would let an AI teach their child or diagnose their illnesseducation policymakerPublic attitudes towards AI in education differ by region and need to be taken into account when introducing it.edtech companyParental trust varies across markets, so localised messaging is important.