Issue 11
Viewpoint
child safety · United States
Eight US states sue OpenAI over alleged harms from ChatGPT
Eight US states have sued OpenAI, alleging that ChatGPT was designed in a way that can endanger young people. Kansas Attorney General Kris Kobach said the states will not wait for federal action.
Who is affected and how
- parent
- The lawsuit puts ChatGPT's child-protection measures under scrutiny from the states' side, so it is worth reviewing the settings on a child's account.
- education policymaker
- The states' independent action signals that they are filling the gap left by missing federal regulation, which calls for risk assessment in school contracts as well.
- edtech company
- For products that use a chatbot and also reach minors, design decisions can become a matter of legal liability.
AI companion · United States
Character.AI files for bankruptcy as states ask court to protect children's data
Character.AI filed for Chapter 11 bankruptcy protection on Monday to sell its assets under court supervision, while lawsuits and investigations in several states over alleged ties to suicides…
parentThe fate of users' chats is uncertain, so it is worth checking what conversations a child had in the app and whether deletion of the data can be requested.education policymakerThe case shows that if an AI companion service goes bankrupt, minors' data can become an asset, which is a regulatory issue.teacherThe AI companion platform popular among students may shut down or change owners, so it is worth discussing the topic in class.case · AsiaStudent caught using AI smart glasses in Yonsei entrance exam; South Korean universities tighten security
A student was caught using AI smart glasses during the Yonsei University entrance exam, after which South Korean universities tightened security checks. Ahead of the upcoming CSAT, fear of cheating…
teacherExam invigilators must prepare to recognise wearable AI devices and to rework their checking procedures.education policymakerExam rules must be extended to wearable devices, because the traditional ban on phones and watches is no longer enough.university studentAdmission and exam regulations may treat wearable AI devices as prohibited aids.child safety · United StatesProposed class action: OpenAI used children's data to train ChatGPT without parental consent
A proposed class action alleges that OpenAI let minors register for ChatGPT without genuine parental consent and used their conversations for model training. The complaint includes the case of a…
parentCheck whether a child under 13 or a teenager uses ChatGPT, and which data-sharing settings are in force.school leaderFor AI tools recommended to students, the data-handling and consent practices for minors must be clarified in advance.edtech companyUsing children's data for model training creates litigation risk, so a documented parental-consent process is required.regulation · United StatesUS schools and states restrict generative AI in lower grades
New York City schools are banning AI through 8th grade this year, Katy ISD bans generative AI through grade 6, and the Frederick County school board voted 6-1 for a ban covering all PreK–12 students…
school leaderBans differ by grade level, so school policy must be aligned with the relevant district's rules.teacherIn some districts, teachers in the lower grades must prevent the use of generative AI.parentThe district's decision determines whether a child may use AI at school.Hungary
school practiceBosch launches national AI education programme in 20 Hungarian secondary schoolsschool leaderThe programme is available through school applications, so it is worth assessing whether the institution can take part.teacherThe company's engineering experts deliver the AI training, so the content of classroom use can be aligned with the programme.studentYou can gain practical skills in phrasing questions to AI and checking the sources of its answers.caseWho decides what students learn: the national curriculum or the algorithm?education policymakerThe requirements of the national core curriculum (NAT) and the actual behaviour of the AI tool a student uses may diverge, so the tools' pedagogical settings should also be examined.teacherIt is worth checking whether the tool students use gives an explanation or only a ready-made solution.school practiceObuda University's AI ambassadors guide others through the shift to AIuniversity studentAI ambassadors are available among fellow students, and you can turn to them with questions about using the tools.education policymakerÓbuda University's model is a template that higher education institutions can adopt for their own internal AI rollout.surveyAt one Hungarian secondary school, more than 90% of students use AI dailyteacherBecause daily use is so widespread, lessons should teach how to use AI rather than dispute whether it is used.school leaderThe school's AI policy should be sized for near-universal student use.regulationSlovak education ministry gives schoolteachers free AI toolsteacherTeachers in Slovakia get access to the main AI tools, which also raises questions about teacher training and rules of use.education policymakerThe neighbouring country's example shows a state-level, multi-provider access model.school practiceTeacher in Szandaszőlős wins FITT award for teaching with artificial intelligenceteacherA Hungarian example shows that a method building AI into teaching can earn professional recognition.studentLocal students' learning is supported by AI through the teacher's method.caseGen Z now asks ChatGPT about intimacy rather than their parents, Hungarian article saysparentTeenagers may bring sensitive questions to a chatbot too, so it is worth keeping the conversation open.teacherThe link between sex education and AI use may also come up in school information.school practiceBudapest conference for educators on AI use in kindergartens and schoolsteacherThe conference offers a chance to learn about practical questions of AI use in kindergartens and schools.school leaderKnowledge gained at the event can be built into the institution's AI guidelines.
More key stories of the week
regulation · European Union
EU AI Act: which school and university AI counts as high-risk
The analysis sets out which school and university AI falls into the high-risk category under point 3 of Annex III of the AI Act. It covers the ban on emotion recognition in class and exams, deployer…
school leaderAs a deployer, the institution has obligations and must therefore assess the systems it uses.edtech companyHigh-risk education systems and prohibited emotion recognition carry separate compliance requirements.education policymakerAligning member-state enforcement with school use requires following the AI Act deadlines.research · globalStudies: ChatGPT homework help can lower exam scores, while a hint-giving AI tutor did not
In a study of nearly 1,000 students, ChatGPT lowered exam scores by 17 percent, while an AI tutor that gave only hints did not. Another report says students who leave their work to AI may get worse…
teacherA hint-giving AI tutor and a chatbot that gives ready answers have different effects, so assignments should be set accordingly.studentIf you ask AI for the finished solution, you can expect worse exam results.parentWhen helping with homework, a tool that gives hints is safer than a chatbot that gives the solution.tool · globalChatGPT for Teens still does homework and raises safety concerns, tests find
Several analyses have raised safety and homework concerns about ChatGPT for Teens, and a nonprofit's tests found that the teen mode still completes homework assignments. One author who held…
parentTurning on teen mode does not guarantee that a child will not be given a ready-made answer, so it is worth reviewing their use together from time to time.teacherAssignments should be designed so that asking for a ready-made answer does not solve them.studentAccording to the tests, teen mode does not stop the system from doing the task for you, which undermines your own learning.tool · globalChatGPT adds flashcards, a college planner and Study Mode for teens and teachers
OpenAI has added new learning tools to ChatGPT for Teens, including flashcards and a planner that helps US students choose a college. Hungarian news reports say a Socratic Study Mode and a course for…
studentFlashcards and Study Mode guide you with questions, but they help only if you supply the answer, not the machine.teacherOpenAI's new course and tool can be tried out, but the rules for classroom use must be set by the teacher.parentThe learning features your child uses differ from a plain chatbot, so it is worth getting to know them.research · globalAI detectors can falsely accuse students: what the false-positive data show
Three overviews discuss the accuracy of AI detectors and their false positives. They say that at scale even a small error rate affects many honest students.
teacherA detector's flag is not proof in itself; suspicion should be checked by other means.studentIf falsely accused, documents of your working process, such as drafts and versions, can offer protection.school leaderThe disciplinary code should set out how false detector flags are handled.survey · United StatesMost New York school districts have no AI policy, surveys find
According to a survey and a Route Fifty investigation, most New York school districts do not have an AI policy. The study tried to collect the rules of dozens of districts.
school leaderIn districts without a policy, the principal decides what can be done with AI, so a written guideline is needed.education policymakerThe scale of the regulatory gap shows that a central template is needed for school guidelines.survey64% of secondary school teachers say AI grading of written work only increased their workload
According to a report, 64% of secondary school teachers say that AI grading of written assessments only increased their workload. The result contrasts with the time savings AI grading promises.
teacherBefore AI grading is introduced, it should be measured how much checking and correction work it creates for teachers.school leaderBefore introducing a grading-support tool, it is worth assessing the time teachers spend on it.edtech companyTeacher surveys call into question the time-saving promises made for products.caseUS charity plans to use AI to monitor Gaza classrooms for 'hateful' speech
A US charity plans to use AI to check classroom speech in its Gaza schools, because it is concerned that teachers deviate from the expected content. The plan tests the limits of teacher monitoring…
teacherAI monitoring of classroom speech affects teacher autonomy and privacy.education policymakerThe case shows that content-monitoring AI requires professional and ethical regulation.AI companion · AsiaSouth Korean lawmakers flag youth protection gaps in AI character chat services
South Korean lawmakers are criticizing gaps in youth protection in AI character chat services, and Scatter Lab has limited young people's use of Zeta AI to three hours a day. According to a report on…
parentThe risks of AI character chats are concrete for teenagers, so both usage time and the content of conversations deserve attention.education policymakerAlongside providers' daily time limits, regulatory obligations must also be defined.edtech companyFor chatbots used by young people, usage limits and handling of risky conversations are expected.researchAustralian education research body urges schools to use AI sparingly
An Australian education research body is urging schools to use AI sparingly because of learning risks. According to AERO, there is little robust evidence that AI has an impact on students' learning.
school leaderIt is advisable to begin AI adoption with evidence-based, narrowly scoped pilots.teacherAI should be used only in a targeted way and for a measurable learning goal.education policymakerGiven the lack of evidence, an impact assessment is needed before national rollout.researchCanadian op-ed says AI chatbots have set back teens' learning and calls on governments to act
A Canadian opinion piece, citing new data, says teens who use AI daily are years behind in their learning. The author argues that governments cannot wait to regulate.
education policymakerData pointing to learning gaps adds urgency to calls for regulatory action.parentBecause of the learning gap among teens who use AI daily, it is worth discussing how they use it.case · AsiaParents urge police action over alleged explicit deepfakes at Seremban school
Parents are urging police action over alleged explicit deepfakes at a school in Seremban, Malaysia. They have told police that the student victims should not have had to face the suspect.
school leaderIn a deepfake case, the handling of reports and the protection of victims must be planned in advance.parentIt is advisable to file both a police report and a report with the school.studentAs a deepfake victim, preserve the evidence and ask for help.In brief
caseTwo teachers jailed over AI-generated nude images and child abuse materialschool leaderTeacher background checks and rules on taking photographs need to be tightened.parentThe risk of school photos and children's pictures being misused with AI is real.education policymakerChild protection rules need to be updated to address deepfakes.child safetyExperts urge caution over uploading children's photos to AI; safety of AI toys questionedparentDo not upload a child's photo to an AI tool without knowing how the data is handled, and check AI toys before buying them.regulationCalifornia enacts companion chatbot child safety requirements and new school AI lawsschool leaderSchools in California should expect an expansion of digital wellness education.edtech companyThe child safety requirements for companion chatbots affect how the product operates.education policymakerCalifornia's rules could serve as a model for other jurisdictions.caseCSUMB students protest California State University's ChatGPT contract with OpenAIuniversity studentStudents can organise to oppose institutional AI contracts.education policymakerA lack of student involvement in AI contracts can provoke controversy.researchVirginia Tech study: students mostly ask ChatGPT to write the essay itselfteacherMaking usage visible may be more effective than detection in curbing requests for ready-made text.university studentIf your workflow is visible, you are less likely to hand the whole essay over to the machine.caseCambridge and Newcastle change their Turnitin deals over AI concernsuniversity studentYou can check the contract terms on how your submitted work is used.edtech companyUniversities are also renegotiating the terms of data use.toolArden University gives students Gemini Enterprise access; IIT Kharagpur launches AI tutor and free GATE-JAM platformuniversity studentSome universities now offer institutional AI access, which may come with enterprise-grade data protection terms.edtech companyAgreements between universities and major cloud providers are narrowing the market.school practiceGoogle offers free AI training to teachers in India and the PhilippinesteacherFree training from a platform provider is available for using AI tools.education policymakerCoordinating public and corporate training programmes could speed up teacher training.researchMicrosoft Research's StudentSim builds one simulated student per learner to improve AI tutorsedtech companySimulated students allow AI tutors to be tested and fine-tuned without real learners.teacherThe work is still at the research stage, so it does not immediately affect classroom use.surveyParents limit children's social media and AI use amid growing cautionparentOther parents' experiences of limiting use offer a model to follow.teacherWith parents growing more cautious, teachers can expect questions about AI use.toolMoms for Liberty uses AI to monitor decisions in 13,000 school districtsschool leaderSchool board decisions may become public before the school itself has communicated about them.parentThe system promises advance alerts about controversial policies.researchUS report: federal funding needed to research AI's impact in schoolseducation policymakerBecause research funding is lacking, decisions on AI adoption are being made without evidence.school leaderIn the absence of research, it is worth documenting the results of your own pilots.regulationPortugal opens public consultation on AI use in schoolseducation policymakerSchool AI guidelines can also be developed through public consultation.school leaderThe flexible Portuguese guidelines can serve as a model for institutional rules.toolUK ed-tech Sainaptic raises £1M for AI-assisted marking; separate AI tutoring platform backed by $8M roundedtech companyInvestors are showing interest in AI-based marking and tutoring.teacherAI-assisted marking may take over some teachers' work, but teachers must still check the results.regulationGovernment directs 80,000 teachers to complete an AI course by the end of the monthteacherTeachers must complete the AI course by the deadline.education policymakerMandatory training is a fast model but places a heavy organisational burden.toolHong Kong police launch AI-powered national security platform for school pupilsteacherA state AI platform can be introduced into the school curriculum.studentStudents may encounter a state-run AI tool in national security education.toolAI weapons detection systems now operating at Hillsboro schoolsschool leaderWhen introducing AI weapons detection, the handling of false alarms and data protection must be clarified in advance.parentThe school uses surveillance AI, and it is worth asking the institution for information about it.caseDutch AI-fraud hearing and AI interviews for kindergarten teacher applicantsuniversity studentUnderstanding of the work may have to be demonstrated orally if AI fraud is suspected.teacherEarly-career teachers may also face AI interviews when applying for jobs.caseThe Chronicle: professors monitor students in the fight against AI as never beforeuniversity studentMonitoring tools may extend to observing students' working process.teacherRethinking assessment methods may help in weighing surveillance against trust.regulationEU consultation invites public views on AI companions and teensparentThe consultation offers a direct way to comment on the regulation of AI companions.education policymakerThe outcome of the consultation could form the basis of EU-level rules on AI companions.regulationUAE makes AI literacy compulsory across K–12 schooling, including kindergartnerseducation policymakerThe Emirati model points to a compulsory-teaching approach, the opposite of a ban.teacherAI literacy becoming a compulsory part of the curriculum requires new teacher preparation.school practiceKenya and Zambia plan to bring AI into education and career guidanceeducation policymakerIn developing countries, both curricula and career guidance are being extended with AI tools.studentA career guidance platform can offer personalised advice.