AI Threat Watch

AI incidents in education and among young people

Verified cases from around the world in which AI harmed students, children, teenagers or the people and institutions that teach them, with statistics, case studies, how schools and authorities responded, and the prevention measures and laws that governments, regulators and education bodies have put in place.

Child protection: no child victim and no child perpetrator is named on this site. A young person is named only when their family made the name public in court or in mainstream media.

95Case files
93Involve a young person
46Happened in a school or university
30Countries
71%Rated critical or high

Cases per year, by where they happened

SchoolUniversityOnlineAt home
010203040School: 8University: 2Online: 4142023School: 12University: 1Online: 5At home: 2202024School: 10University: 4Online: 13At home: 7342025School: 7University: 2Online: 14At home: 4272026

Each case is counted once, by the year it was first reported. Where it happened is kept separate from who was involved: a teenager harmed by a chatbot at home counts as a young person, not as a school incident.

Types of incident

Deepfake abuse43
Chatbot harm35
AI misfire6
AI exam fraud5
AI-assisted malware2
Model & data leak2
Deepfake fraud1
Prompt injection1

Where it happened

School37
University9
Online36
At home13

Age of the young people involved

Children (under 13)4
Teenagers (13 to 17)73
Young adults (18 to 24)14
Age not stated2

Where in the world

CountryCase files
United States42
South Korea6
Australia5
United Kingdom5
Canada3
Japan3
India3
Argentina2
Brazil2
Spain2
Italy2
Mexico1

Countries with the most case files. Counts reflect where incidents are reported and researched, so they understate countries with less media coverage in the languages we monitor.

Case studies

Selected cases, with what happened and how schools and authorities responded, as recorded in each case file.

Deepfake abuseSouth Korea

South Korea school deepfake crisis: Telegram rooms target students; President Yoon orders crackdown

What happened
In August 2024 Telegram chatrooms tied to schools and universities were found sharing AI sexual deepfakes of female students, including one with about 1,200 members targeting more than 30 Inha University students; police said 73.6% of suspects identified in early 2024 were minors. President Yoon ordered a crackdown, police began a seven-month special operation, and on 26 September parliament made possessing or watching sexual deepfakes a crime.
How schools and authorities responded
Seven-month KNPA crackdown with detection software; National Assembly criminalised buying, saving or watching sexual deepfakes (up to 3 years) and raised the maximum for making them to 7 years
Status
Ongoing

Deepfake abuseSpain

AI 'undress' app used to make fake nudes of 20+ schoolgirls in Almendralejo; 15 minors later convicted

What happened
Boys aged 12 to 14 took girls' photos from social media, used an AI app to put their faces on naked bodies and shared the images in WhatsApp and Telegram groups, prompting at least 22 police complaints. In July 2024 the Badajoz juvenile court convicted 15 minors on 20 counts of creating child sexual abuse material and 20 counts against moral integrity, giving each a year of probation.
How schools and authorities responded
National Police and the Central Cybercrime Unit investigated, the case went to the juvenile prosecutor and Spain's data protection agency opened its own inquiry; 15 boys got one year of probation plus classes on gender equality and responsible technology use
Status
Resolved

Deepfake abuseUnited States

Athletic director frames Pikesville High principal with AI-generated racist and antisemitic audio

What happened
An audio clip appearing to show Pikesville High principal Eric Eiswert making racist and antisemitic remarks spread on social media in January 2024; police and an FBI-contracted analyst found it was AI-generated with human editing. Former athletic director Dazhon Darien was arrested in April 2024 and in April 2025 took an Alford plea to disturbing school operations, receiving four months in jail.
How schools and authorities responded
Arrest on 25 April 2024 on charges including stalking, theft, retaliation against a witness and disrupting school operations; Alford plea in April 2025 with other charges dropped; Eiswert sued the school district and Darien in January 2025.
Status
Resolved

AI misfireUnited States

Omnilert AI gun detection flags a student's chip bag at a Baltimore County school; police handcuff him

What happened
Omnilert's AI camera system flagged an empty Doritos bag held by a student after football practice as a possible firearm, and armed police searched and handcuffed him. The district's security team had already cancelled the alert, but the principal, unaware of the cancellation, reported it to the school resource officer, who called police.
How schools and authorities responded
Omnilert expressed regret but said the process functioned as intended; county council members called for a review of the system; the superintendent said reviews are part of regular practice
Status
Resolved

Chatbot harmLebanon

Lebanese family sues OpenAI, alleging ChatGPT contributed to their 15-year-old son's suicide

What happened
The parents of Matias Habib, who died aged 15 in April 2026, filed a wrongful-death suit against OpenAI and Sam Altman in San Francisco, alleging ChatGPT steered him towards nihilism and failed to act when he asked about suicide.
How schools and authorities responded
Suit filed in the Superior Court of California, County of San Francisco, by the Social Media Victims Law Center and the Tech Justice Law Project, within the coordinated ChatGPT proceeding (JCCP). OpenAI had not commented publicly at the time of reporting.
Status
Ongoing

AI exam fraudGH

WAEC withholds WASSCE results for candidates at 235 Ghanaian schools over AI-generated answers

What happened
The West African Examinations Council withheld subject results of candidates from 235 schools in Ghana for collusion and use of AI-generated answers in the 2023 WASSCE. WAEC said some candidates copied chatbot error replies such as "I'm not familiar with the term you have used" into their answer booklets.
How schools and authorities responded
Results withheld pending investigation; WAEC said school staff and parent complicity could not be ruled out
Status
Under investigation

Model & data leakUnited States

AP/Seattle Times obtain ~3,500 unredacted student files flagged by Gaggle AI monitoring in Vancouver, WA

What happened
A records request by The Seattle Times and AP returned links to nearly 3,500 unredacted screenshots of student writing flagged by Gaggle's AI surveillance, covering suicide, depression, abuse and sexuality, viewable by anyone with the link and without a password. At least six students were potentially outed to school officials over writing about being gay or transgender.
How schools and authorities responded
District apologised; Gaggle changed the system so screenshots require login after 72 hours
Status
Contained

How schools and authorities responded

The response recorded in each of the 46 case files that happened in a school or university, from the sources cited in the case file.

CaseHow schools and authorities responded
Kochi college student arrested for AI-morphed sexual images of teachers shared in Telegram groups India, 2026College expulsion; arrest and remand under the BNS, IT Act sections 67 and 67A and the Kerala Police Act; special investigation team with cyber experts.
Tanzania's NECTA cancels 58 Form Six results after candidates used phones to access ChatGPT in exams TZ, 2026Results cancelled; NECTA said AI tools cannot give candidates an advantage
New Zealand teachers left unable to work after students make AI deepfake porn and TikTok fakes of them New Zealand, 2026Videos reported to TikTok and Netsafe; content removable under the Harmful Digital Communications Act.
Guwahati schools suspend students over AI deepfake images of female students and teachers India, 2026Matter referred to the school's Child Abuse Monitoring Committee; students suspended; meetings with parents.
Louisiana student charged with terrorizing after AI-made gun photo forces Southside High shelter-in-place United States, 2026Student charged under La. R.S. 14:40.1 (terrorizing); police chief announced zero tolerance for such threats.
FSU shooting victims' families allege ChatGPT aided the 2025 campus gunman United States, 2026OpenAI said ChatGPT gave factual responses with information broadly available online and that it cooperated with law enforcement. It remains unclear how far the responses influenced the attack.
Tumbler Ridge shooting victims sue OpenAI over unreported flagged ChatGPT account Canada, 2026Sam Altman publicly apologised for not alerting police to the banned account. OpenAI says it reports imminent, credible threats and has strengthened safeguards.
Heraklion pupils use AI to turn photos of female teachers and classmates into pornographic images GR, 2026School administration acted on the complaint, with child psychologists helping manage the case; no police action was reported.
PSNI investigates AI-generated explicit images of pupils shared at Royal School Armagh United Kingdom, 2026School referred the matter to authorities; PSNI investigating with the school and families; UK government introduced legislation to criminalise creating sexual deepfakes without consent
Incheon high school student charged, then jailed, for sexual deepfakes of two teachers and classmates South Korea, 2025Charged under the sexual crimes special act; sentenced on 16 September 2026 and taken into custody, with a treatment program and a five-year ban on youth-related jobs; Incheon education office pledged legal and psychological support for the teachers.
Louisiana boys charged under new AI law over nude images of 8 girls; a victim had been expelled United States, 2025Sheriff charged one boy (booked 15 September 2025) and a second in December with 10 counts each; the school board reinstated the girl on probation; the district said it was starting to develop AI policies
Dozens of Yonsei University students admit using ChatGPT in an online midterm for a ChatGPT course South Korea, 2025Yonsei planned an emergency public hearing via its Institute for AI and Social Innovation
Omnilert AI gun detection flags a student's chip bag at a Baltimore County school; police handcuff him United States, 2025Omnilert expressed regret but said the process functioned as intended; county council members called for a review of the system; the superintendent said reviews are part of regular practice
Students use AI and a classmate's photo to fake a weapons threat against a Michigan high school United States, 2025Increased police presence on 20 October; students removed from school; case referred to the county sheriff for charging review.
Australian Catholic University logs ~6,000 misconduct cases, mostly AI, often relying on Turnitin's detector Australia, 2025ACU switched off the Turnitin AI indicator in March 2025, expressed regret and changed its integrity process
Japanese police: classmates made over half of sexual deepfakes of minors, many from school photos Japan, 2025NPA studied about 10 popular image-generating apps; police used defamation and obscenity laws; Tottori Prefecture banned making sexual fake images of children in August 2025
Nine Lawrence, Kansas students sue district over Gaggle AI surveillance that flagged and removed their work United States, 2025District declined to comment on litigation; it had said in 2024 it would keep using Gaggle under a three-year contract ending July 2026
HKU law student accused of making 700+ AI porn images of 20-30 women, including classmates and teachers Hong Kong, 2025HKU issued a warning letter and required an apology, then said the case remained open; the Office of the Privacy Commissioner for Personal Data began a criminal investigation; the government said it would review the law if needed.
Nikkei finds hidden prompts telling AI reviewers to praise papers in 17 preprints from 14 universities Japan, 2025KAIST disowned the practice; publishers' policies on AI review diverge
Vietnam police open criminal cases after candidates used ChatGPT, Gemini and StudyX in national exam Vietnam, 2025Criminal cases under Penal Code articles on state secrets; Ministry of Public Security ordered urgent investigation
Adelaide private school suspends student over a deepfake image of a teacher; police informed Australia, 2025Student suspended indefinitely; SA Police informed. The case came days after South Australia passed a law criminalising explicit or violent deepfakes.
AP/Seattle Times obtain ~3,500 unredacted student files flagged by Gaggle AI monitoring in Vancouver, WA United States, 2025District apologised; Gaggle changed the system so screenshots require login after 72 hours
Omnilert AI gun detection at Nashville's Antioch High fails to spot the gun in a fatal school shooting United States, 2025District said no single system catches everything; a wounded student sued Omnilert and installer System Integrations in May 2026; the district later installed Evolv scanners in middle and high schools
Two Lancaster Country Day School boys charged over AI-altered sexual images of 59 minors United States, 2024The DA filed juvenile charges and found no criminal liability for school staff under existing reporting law; Pennsylvania's new AI child abuse material law took effect on 20 December 2024; both boys got probation, 60 hours of community service and no-contact orders
FTC acts against Evolv over AI weapons scanners that missed a knife used to stab a student at a school United States, 2024Proposed order lets K-12 customers who signed April 2022 to June 2023 cancel; no monetary relief
Singapore police probe deepfake nudes of Singapore Sports School girls made by male peers Singapore, 2024School lodged a police report, caned some students and imposed suspensions and sports trip bans; police interviewed students, seized devices and requested removal of images
Seoul court jails SNU graduate for 10 years over Telegram deepfakes of at least 61 women South Korea, 2024Defendants arrested and convicted; Park got 10 years and Kang four years in the first trial
Two Class IX students booked in Moradabad, India for posting an AI-made obscene image of their teacher India, 2024FIR under the IT Act; police investigation and takedown efforts.
South Korean teachers' union finds 204 teachers hit by school deepfake sex crimes in Telegram scandal South Korea, 2024The Ministry of Education set up an emergency deepfake task force; ministries agreed to build automatic detection of deepfake sexual content; police opened an investigation into Telegram.
Guatemalan prosecutors raid homes over students' AI-altered sexual images of schoolgirls GT, 2024Complaint filed 10 August 2024; raids and device seizures; SVET filed a complaint on the Vice President's order; Mineduc applied a protocol and the school separated the suspects and reported to the National Civil Police
Whistleblower says LAUSD's $6M 'Ed' AI chatbot sent student data offshore before vendor AllHere collapsed United States, 2024LAUSD dropped AllHere and said Ed data stays district-owned and protected; inspector general interviewed the whistleblower
Melbourne student expelled for making sexually explicit AI images of a female teacher at Salesian College Australia, 2024Student expelled; school says it introduced cyber safety and respectful relationships programs.
Turkish police jail university entrance exam candidate who used a button camera and AI to get answers Turkey, 2024Arrests; police released a video demonstrating the AI device
Chilean prosecutors open probe after six students spread AI nude images of seven classmates Chile, 2024Prosecutors opened a probe on their own initiative; the school applied conditional enrollment and other sanctions but did not expel the boys; parents sought protection orders in the Court of Appeals
Beverly Hills district expels five eighth-graders over AI nude images of 16 classmates United States, 2024The district investigated from February 2024 and its board expelled five students; Beverly Hills police opened a criminal investigation with no arrests reported
Athletic director frames Pikesville High principal with AI-generated racist and antisemitic audio United States, 2024Arrest on 25 April 2024 on charges including stalking, theft, retaliation against a witness and disrupting school operations; Alford plea in April 2025 with other charges dropped; Eiswert sued the school district and Darien in January 2025.
WAEC withholds WASSCE results for candidates at 235 Ghanaian schools over AI-generated answers GH, 2023Results withheld pending investigation; WAEC said school staff and parent complicity could not be ruled out
Two Miami middle-schoolers, 13 and 14, charged with felonies over AI nude images of classmates United States, 2023School suspended both boys for 10 days and reported to police; Miami-Dade police charged them with third-degree felonies under Florida's 2022 altered sexual depiction law
Rio police investigate AI fake nudes of about 20 girls made by classmates at Barra da Tijuca school Brazil, 2023DPCA and the 16th Police Precinct opened an inquiry and identified some students; the school called the events 'regrettable' and promised disciplinary measures
Westfield High School students allegedly used AI to make pornographic images of female classmates United States, 2023Principal emailed parents, school investigated with the school resource officer and Westfield police and offered counselling; no public discipline or charges
IPN expels student accused of AI-altering classmates' photos into sexual content; later jailed Mexico, 2023IPN expulsion; arrest by Mexico City prosecutors; partial acquittal in December 2024, which the victims' lawyer appealed; five-year sentence and a fine of 103,740 pesos for storing child sexual abuse material
Ecuador prosecutors probe AI sexual montages of 20 to 24 schoolgirls at a private Quito school EC, 2023Prosecutors opened an investigation into child pornography; the Education Ministry cited its new digital violence protocol of 28 September 2023; the two suspects left the school
AI 'undress' app used to make fake nudes of 20+ schoolgirls in Almendralejo; 15 minors later convicted Spain, 2023National Police and the Central Cybercrime Unit investigated, the case went to the juvenile prosecutor and Spain's data protection agency opened its own inquiry; 15 boys got one year of probation plus classes on gender equality and responsible technology use
Peru prosecutors probe two Lima schoolboys accused of selling AI-made nudes of female classmates PE, 2023Family prosecutors opened an investigation and visited the school; the school suspended classes for a day and moved the suspects to virtual classes; the Ministry of Women demanded punishment
Texas A&M-Commerce instructor asks ChatGPT if it wrote essays, then withholds grades of flagged seniors United States, 2023University investigated case by case, cleared several students and began developing AI policies
High school students post AI deepfake TikToks of a Carmel, New York principal making racist threats United States, 2023The Board of Education condemned the videos; the students were disciplined, including suspensions. No criminal charges were reported.

Prevention measures in place

Measures that ministries, regulators, exam bodies and international organisations have adopted or published, with what each one does, as stated in the document.

Ministries and governments

  • UK Department for Education · England · 2025Generative artificial intelligence (AI) in educationDfE policy (updated August 2025) says safety should be the top priority, pupils should use GenAI only with safeguards such as supervision and filtering and monitoring, and personal data should not be put into GenAI tools. It warns GenAI can make cyber attacks and impersonation emails more credible and points to JCQ and Ofqual rules on AI malpractice.
  • UK Department for Education · England · 2026Generative AI: product safety standardsStandards for edtech suppliers (first issued January 2025, expanded 19 January 2026) covering filtering, monitoring and reporting to the safeguarding lead, security against jailbreaking, privacy, testing and governance. The 2026 update adds cognitive development, emotional and social development, mental health and anti-manipulation standards, including avoiding anthropomorphism and emotional dependence and detecting distress.
  • UK Department for Education · England · 2026Cyber security standards for schools and collegesSets standards including annual cyber risk assessment, staff and student training, MFA for staff cloud/remote and admin accounts, patching critical vulnerabilities within 14 days, and a 3-2-1 style backup plan. It also requires reporting cyber attacks internally and to bodies such as insurers and the DfE (last updated September 2026).
  • UK Council for Internet Safety / DfE · England · 2024Sharing nudes and semi-nudes: advice for education settings working with children and young peopleThe March 2024 update explicitly covers AI-generated and digitally manipulated 'pseudo-images' of under-18s and says the law applies to children who make them. It tells staff to refer to the designated safeguarding lead, avoid viewing, copying or sharing images, refer to police or social care where needed, and use removal tools such as IWF/Childline Report Remove.
  • UK Department for Education · England · 2026Keeping children safe in education 2026Statutory safeguarding guidance for schools and colleges, with the 2026 version in force from 1 September 2026. DfE's GenAI guidance names it as the core reference for online safety, filtering and monitoring and cyber security when schools use AI.
  • New York State Education Department · New York State, USA · 2023Determination on biometric identifying technology in schoolsThe September 2023 determination bars New York schools from buying or using facial recognition technology, after a state report found the risks outweighed benefits, including higher misidentification rates for children and other groups. Other biometric tools are left to local decisions that must weigh privacy, civil rights, effectiveness and parental input.
  • Republic of Korea Ministry of Education (reported by The Korea Herald) · South Korea · 2024Education Ministry response to school deepfake sex crimesIn August 2024 the ministry set up an emergency task force to track school deepfake cases weekly, handle victim reports, give psychological support and strengthen digital-ethics education. It said high-school perpetrators could face expulsion, the top sanction in the school-violence system.
  • Republic of Korea Ministry of Education (reported by The Korea Times) · South Korea · 2024799 students, 31 teachers victimized by deepfake videos this yearMinistry survey data showed 833 school victims of deepfakes in 2024 to late October, including 799 students and 31 teachers. Schools referred 417 cases for investigation and 218 for deletion of videos.
  • Ireland Department of Education and Youth · Ireland · 2025Guidance on Artificial Intelligence in SchoolsNational guidance published 21 October 2025 with Oide to help schools plan safe, ethical and appropriate AI use, drawing on UNESCO and EU work. It notes the State Examinations Commission requires AI-generated material to be referenced, and a taskforce on AI in education including assessment was announced.
  • UAE Ministry of Education (reported by Education Middle East) · United Arab Emirates · 2026Safe and Responsible Use of Artificial Intelligence in Classrooms (manual)The 2026 manual bans generative AI for students under 13 or below Year 7, bans AI in formal exams and undisclosed AI work, and prohibits entering personal data of students, teachers or parents into AI tools. It also prohibits deepfakes and impersonation and requires direct teacher oversight of classroom AI use.
  • Saudi Ministry of Education and SDAIA (reported by Gulf News) · Saudi Arabia · 2025Guidelines for using generative AI in educationJoint guidebook for students, teachers and parents aimed at ethical and responsible use that keeps the teacher central. It says students under 13 should not use GenAI without direct supervision and under-18s need clear parental consent.
  • Center for Educational Research and Development (CRDP), Lebanon · Lebanon · 2025Lebanon's AI Competency Framework / AI Educational Framework for students and teachersCRDP, which develops Lebanon's national curriculum, published (October 2025) an AI framework whose first part is an AI policy for Lebanese schools built on UNESCO's 2024 frameworks and 2021 AI ethics recommendation. Its guiding principles include a human-centred approach, data privacy and security, inclusivity and teacher professional development.
  • Egypt Ministry of Education and Technical Education (reported by Egyptian Streets) · Egypt · 2025AI curricula in schools from 2025-2026The minister announced AI curricula in selected primary, preparatory and secondary grades from 2025-26, including the ethics and social implications of AI. No published usage or safeguarding guideline from the ministry was found.

Regulators and law enforcement

  • European Union · European Union · 2024Regulation (EU) 2024/1689 (AI Act), Annex III point 3 (education)Classifies as high-risk AI used to decide admission or assignment to education, to evaluate learning outcomes, to assess the level of education a person will receive, and to monitor and detect prohibited behaviour by students during tests. High-risk systems carry obligations such as risk management, human oversight and documentation.
  • UK Information Commissioner's Office · United Kingdom · 2024Essex school reprimanded after using facial recognition technology for canteen paymentsThe ICO reprimanded Chelmer Valley High School (23 July 2024) for introducing facial recognition without a data protection impact assessment, without consulting its DPO, parents or pupils, and relying on opt-out consent. It stresses a DPIA is required before biometric systems and that explicit opt-in consent is needed.
  • US Cybersecurity and Infrastructure Security Agency (CISA) · United States (K-12) · 2023Protecting Our Future: Partnering to Safeguard K-12 Organizations from Cybersecurity ThreatsCISA's January 2023 report tells schools to start with the most impactful measures: MFA, patching known exploited vulnerabilities, tested backups, an exercised incident response plan and ongoing training. It also urges pressing vendors for secure-by-default products and joining information-sharing groups such as MS-ISAC and K12 SIX.
  • US Federal Bureau of Investigation (IC3) · United States · 2025Senior US Officials Impersonated in Malicious Messaging Campaign (PSA I-051525-PSA)Warns that criminals are using AI-generated voice messages and texts to impersonate senior officials and steal account access. It advises verifying identity through a known number, not clicking links until verified, enabling two-factor authentication, using agreed code words, and confirming payment requests through a separate channel.
  • US Federal Trade Commission · United States · 2025FTC launches inquiry into AI chatbots acting as companionsOn 11 September 2025 the FTC ordered seven firms, including Alphabet, Character.AI, Meta, OpenAI, Snap and xAI, to explain how they test, monitor and limit harms of companion chatbots to children. It asks about monetising engagement, age-restriction enforcement, disclosures to parents and use of conversation data.
  • eSafety Commissioner (Australia) · Australia · 2026Toolkit for Schools: Guide to responding to image-based abuse involving AI deepfakesStep-by-step guide (first released mid-2025, current PDF dated March 2026) for school leaders responding to nude or sexual deepfakes of students or staff, designed to sit alongside child-protection policies. eSafety urges schools to prioritise the wellbeing of targeted students and staff, report possible crimes to police, meet mandatory reporting duties and use eSafety for rapid removal.
  • eSafety Commissioner (Australia) · Australia · 2025Online safety advisory: AI chatbots and companions - risks to children and young peopleThe 18 February 2025 advisory warns that AI companions simulate personal relationships, allow unregulated discussion of sex, self-harm and suicide, and often lack effective age checks. eSafety later issued legal notices (October 2025) to Character.AI, Nomi, Chai and Chub AI and reported that most companions failed to refer users to support for suicide or self-harm.

Exam and quality bodies

  • Joint Council for Qualifications (JCQ) · UK (GCSE, A level and other general qualifications) · 2025AI Use in Assessments: Your role in protecting the integrity of qualificationsGuidance effective 30 April 2025 treats AI misuse that leaves work not the student's own as malpractice and requires centres' malpractice policies to cover AI. It recommends supervised in-class work, checking drafts, short viva-style discussions and restricting AI on centre devices, says detectors must not be relied on alone, and says AI cannot be the sole marker.
  • Quality Assurance Agency for Higher Education (QAA) · UK higher education · 2023Maintaining quality and standards in the ChatGPT era: advice on generative AI toolsQAA advises universities that managed use of generative AI is preferable to outright bans, while maintaining academic standards. It offers a four-step process for identifying assessments that need adjusting and urges robust internal quality assurance.
  • Tertiary Education Quality and Standards Agency (TEQSA) · Australia (higher education) · 2025Enacting assessment reform in a time of artificial intelligenceNon-prescriptive TEQSA resource (released September 2025) that helps universities manage generative AI risks to the assurance of learning while teaching ethical use. It maps the main strategic pathways institutions are taking to redesign assessment.

International organisations

  • UNESCO · Global · 2023Guidance for generative AI in education and researchUNESCO's first global guidance on generative AI in education (published September 2023, page updated January 2026) asks governments to regulate GenAI tools and build coherent policy frameworks for education and research. It proposes an age limit for independent conversations with GenAI platforms, mandatory data-privacy protection, and an ethical validation process for tools because most schools cannot yet vet them.
  • UNESCO · Global · 2024AI competency framework for studentsSets out 12 competencies across four dimensions (human-centred mindset, ethics of AI, AI techniques and applications, AI system design) at three levels: understand, apply, create. It is meant to help curricula prepare students to engage with AI safely and to exercise critical judgement of AI solutions.
  • UNESCO · Global · 2024AI competency framework for teachersDefines 15 competencies across five dimensions, including ethics of AI, AI pedagogy and AI for professional learning, at three levels: acquire, deepen, create. It is built on protecting teachers' rights and enhancing human agency, and gives strategies for putting ethical principles into practice.
  • UNICEF · Global · 2025Guidance on AI and children, version 3.0The December 2025 update of UNICEF's policy guidance sets ten requirements for child-centred AI, including regulatory oversight, child safety, protection of children's data and privacy, non-discrimination, transparency and accountability, and building children's AI skills. It comes with a checklist for governments and companies.
  • UNICEF · Global · 2026'Deepfake abuse is abuse' (statement)UNICEF's 4 February 2026 statement cites UNICEF/ECPAT/INTERPOL research finding at least 1.2 million children in 11 countries had their images turned into sexual deepfakes in the past year. It calls on governments to bring AI-generated content into child sexual abuse material laws and criminalise its creation, acquisition, possession and sharing, and on AI firms to build in safety by design and prevent circulation.
  • OECD · OECD members · 2026OECD Digital Education Outlook 2026: Exploring effective uses of generative AI in educationFinds that general-purpose GenAI boosts task performance without real learning gains unless guided by pedagogy, and that students' advantage can disappear or reverse in exams once AI access is removed. It recommends purpose-built educational tools, equitable infrastructure and sustained professional learning.
  • Council of Europe · 46 Council of Europe member states · 2026Committee of Ministers recommendation on AI literacyRecommendation to member states to help pupils, students, teachers, parents and citizens understand AI, use it responsibly and recognise where it threatens human rights, democracy and the rule of law. It gives effect to Article 20 of the Council of Europe Framework Convention on AI.
  • European Commission · European Union · 2026Ethical guidelines on the use of AI and data in teaching and learning for educators (updated)First issued in 2022 and updated in 2026 through the European Digital Education Hub, the guidelines explain the GDPR and AI Act context for teachers. They provide guiding questions and case examples for judging AI tools before using them in class.

Laws that apply

Laws in force that criminalise sexual deepfakes of minors or regulate AI used in education or by young people.

CountryLawWhat it does
United StatesTAKE IT DOWN Act (2025)Criminalises publishing non-consensual intimate images including AI deepfakes (up to three years' prison where the victim is a minor) and requires platforms to remove reported material within 48 hours from May 2026.
South KoreaAmendment to the Act on Special Cases Concerning the Punishment of Sexual Crimes (deepfake possession and viewing) (2024)Makes buying, saving, possessing or watching sexually explicit deepfakes punishable by up to three years in prison and raises the maximum for creating them to seven years.
United Kingdom (England and Wales)Data (Use and Access) Act 2025, section 138 (Sexual Offences Act 2003 ss.66E-66H) (2025)Creates offences of creating, or requesting the creation of, sexual deepfakes of adults without consent, in force from 6 February 2026 (images of under-18s are already covered by indecent pseudo-photograph law).
United KingdomCrime and Policing Act 2026 (child sexual abuse image-generators) (2026)Makes it an offence to make, adapt, possess or supply tools such as AI models built to generate child sexual abuse images, with up to five years in prison; the Northern Ireland provision checked (s.73) was not yet in force.
United Kingdom (England)Indecent images of children law as applied in UKCIS guidance (pseudo-photographs) (2024)Official guidance confirms that making, possessing or sharing indecent AI-generated or manipulated images of under-18s is illegal, including when the maker is a child.
AustraliaCriminal Code Amendment (Deepfake Sexual Material) Act 2024 (No. 78, 2024) (2024)Creates a standalone federal offence of sharing sexual material of an adult without consent, expressly including material created or altered with technology such as deepfakes.
FranceCode pénal article 226-8-1 (inserted by Law No. 2024-449 'SREN' of 21 May 2024) (2024)Punishes sharing an AI-generated or edited sexual image or voice of a person without consent by up to two years and EUR 60,000, or three years and EUR 75,000 if shared online, even if labelled as fake.
ItalyLaw No. 132/2025 on artificial intelligence (Penal Code art. 612-quater) (2025)Makes spreading harmful AI-generated or altered images, video or voice without consent a crime punishable by one to five years, and requires parental consent for under-14s to use AI.
European UnionAI Act, Regulation (EU) 2024/1689, Article 5(1)(f) (2024)Prohibits AI systems that infer emotions of people in education institutions and workplaces, except for medical or safety reasons.
United States (California)SB 243 (companion chatbots) (2025)From 1 January 2026 requires companion chatbot operators to disclose they are AI, run suicide and self-harm protocols, give minors break reminders, and block sexually explicit chatbot images for minors.

All 95 case files

How cases are included: a verified case file appears here when a person under 25 was harmed or involved, or when it happened in a school, university or education service. Each case records separately whether a young person was involved, their age group and where it happened. Data: education.json (CC BY 4.0)

Language: EnglishFrançaisEspañolالعربية中文Русский