Issue 12
Eight US states sue OpenAI over alleged ChatGPT risks to children
Eight US states have sued OpenAI, alleging that ChatGPT encouraged self-harm, reinforced delusions, gave dangerous medical advice and helped plan violence. Texas has filed a separate suit alleging that OpenAI and Sam Altman falsely marketed the product as safe; the parties are demanding stronger age verification and parental controls, and a California suit challenges the collection of minors' data.
Who is affected and how
- parent
- Check that parental controls are switched on for your child's ChatGPT account, and talk through the risks of dangerous advice with them.
- education policymaker
- The suits may set a precedent for chatbots' obligations toward minors; take this into account in your own regulatory planning.
- edtech company
- Expect stricter age verification and safeguards for products aimed at minors.
Teen-safety tests: ChatGPT parent alerts triggered in 5% of simulated cases
A RAND–Harvard preprint says that with parental controls switched on, ChatGPT notified a parent in 15 of 300 simulated teen-risk conversations. Other tests measured 18 chatbots on teen safety, and an…
parentDo not rely on parent alerts as the only protection; also monitor conversations and your child's usage habits separately.school leaderWhen choosing tools approved for teenagers, also take independent safety test results into account.edtech companyFor products aimed at minors, safeguards should also be verified in external, standardised tests.regulation · European UnionProposed EU KIDS Act sets rules on age verification and AI chatbots
A Greenberg Traurig analysis says a proposal has been drawn up for EU rules on online platforms, age verification and AI chatbots. The draft aims to protect children online.
education policymakerCompare the planned age-verification and chatbot requirements with national child-protection rules.edtech companyChatbots on the EU market that children can also access need age verification and a compliance plan.researchAustralian evidence body names 'performance paradox': AI improves task results but reduces learning
Australia's AERO finds that students perform better on a task with AI but learn less in the process. Learning-science guides apply the same principle when weighing when AI should be used in class.
teacherAllow AI at stages where it does not replace the effort of learning.school leaderMeasure implementation guidelines by learning outcomes, not by task completion.school practice · United StatesNew York bans student AI use, Los Angeles pauses it, while other districts leave the decision to teachers
According to a parent newsletter, New York, the largest school district in the US, has introduced a ban covering nearly half a million students. Los Angeles is pausing, while schools around…
teacherCheck which AI tools are permitted in class in your district, because bans differ.school leaderThe differing district approaches (ban, pause, teacher discretion) offer a useful benchmark for your own policy.parentAsk the school which tools are allowed and what the rule is for homework.Hungary
More key stories of the week
Teens can access 'nudify' apps despite Apple and Google parental controls, report finds
The Guardian reports that US teens can access nudify apps despite Apple's and Google's parental controls. A Hungarian-language report on the Tech Transparency investigation has also appeared.
parentBuilt-in parental settings are not enough; check which apps are actually accessible, and talk with your child too.school leaderSchools should prepare a written procedure for deepfake cases.edtech companyGaps in app-store vetting raise questions of platform responsibility.school practice · United StatesUS school districts set AI rules: bans, approved tools and teacher discretion
This week Montclair barred student AI use through grade 8, Albany limited it to approved tools, and Clover is blocking most AI tools on student devices. In Maryland, four districts missed the state…
teacherIn several districts the teacher decides how much AI use is allowed on an assignment; this is worth stating clearly in advance.school leaderThe approved-tool list and meeting deadlines are the key points of a policy.parentIn some districts parents are calling for a pause; it is worth checking the school's policy.education policymakerThe missed Maryland deadline shows that implementation of state rules can lag in the districts.tool · globalGoogle launches free AI tool for teachers; Jisc partnership in UK, 96,000 Gemini licences in Ghana
Google has launched a free AI tool for teachers, entered a two-year partnership with the UK's Jisc, and the University of Ghana has secured 96,000 Gemini licences. For education Workspace…
teacherFree Google tools are available, but institutional data protection terms should be checked first.school leaderWorkspace for Education subscribers can switch on Gemini for Education free of charge.edtech companyGoogle's free package puts pressure on paid educational AI providers.survey · Asia82% of Indian students fear AI speeds up work at the cost of deep learning
According to The Times of India, 82% of the Indian students surveyed fear that AI speeds up their work but at the expense of deep learning.
studentStudents themselves sense the risk of superficial learning, so it is worth choosing deliberately how to use AI.teacherAssignments should be designed so that, beyond speeding up work with AI, they also require deep understanding.research · globalStudy finds AI tutor produced more learning in less time than in-class active learning
A study found that students learned significantly more with an AI tutor in less time than with in-class active learning. According to the cited Scientific Reports article, AI speeds up learning when…
teacherA structured AI tutor can complement classroom work, but controlled research conditions are not the same as everyday use.edtech companyDemonstrating learning gains requires product evidence based on credible measurement.case · AsiaSouth Korean middle schooler who made deepfakes of classmates hospitalised after assault by peers
A South Korean middle school student who made and distributed deepfake videos from photos of female classmates was hospitalised after being assaulted by an acquaintance of a victim. The case…
school leaderSchools must prepare in advance to handle deepfake cases, because vigilante retaliation only makes the situation worse.parentTalk with your child about why other people's images must not be manipulated with AI.case · United StatesMother criticises school over AI-generated photo of her 6-year-old daughter aged 25
After a school AI exercise, Nikki Anderson's six-year-old daughter brought home an AI-generated photo of herself at 25, and her mother criticised the school. A Hungarian-language report also covered…
parentAsk whether the school uses image-generation tools and whether it obtained your consent.teacherParents must be informed and give consent before an AI exercise using children's photos.school leaderA clear rule is needed on AI processing of children's images.school practice · globalTeacherless school says AI can teach students in two hours a day; debate over teacher shortage
A CBS News report describes a teacherless school that teaches students with AI for two hours a day. A Fortune opinion piece argues that every child will get access to AI, but not every child will get…
teacherAI-based models raise questions about the future of the teacher's role.education policymakerAddressing teacher shortages with AI carries a risk of inequality.parentIt is worth asking the school how AI supplements the work of teachers.AI companion · United StatesUS teens turn to AI for dating advice and companionship
Several US reports say teenagers are asking AI for relationship advice and companionship. A 2026 Pew survey found that 64% of US teens have used an AI chatbot.
parentTalk openly with your child about what they use the chatbot for, rather than simply banning it.teacherDigital-skills lessons should treat the emotional risks of chatbots as a separate topic.fraud · AsiaAI glasses and online cheating at South Korean universities: more than 600 students to retake exam
Yonsei University is weighing a criminal complaint against a student caught wearing AI glasses during an essay exam, while more than 600 students at Konkuk University must retake an online exam over…
university studentUsing AI glasses in an exam can lead to sanctions, including a criminal complaint.teacherOnline and AI-glasses exams require new methods of supervision.education policymakerExam security rules need to be adapted to new wearable devices.survey · European UnionFrench survey: 40% of 11–12-year-olds use AI weekly, but only 26% of parents know
A French YouGov survey found that 40% of 11–12-year-olds use AI weekly, but only 26% of parents are aware of it. Some 42% of parents do not feel adequately supported.
parentAsk your child about their AI use, because parents' estimates fall well short of reality.school leaderInformation sessions and training for parents could fill a gap.tool · United StatesOpenAI teacher workshops yield reusable AI tools, but teachers want more practice
OpenAI's AI workshops for teachers produced reusable tools, but participants say they want more practice. A $1 million initiative supports teachers in directing the AI tools used in their classrooms.
teacherThe templates and tools built in the workshop can be reused, but real practice takes time.edtech companyTeachers' need for practice points to a direction for product and training development.school leaderAI training for teachers should include time to practise.