Short answer: how should preschools use AI?
In preschools, AI should be a tool for adults: reducing paperwork, translating and drafting notices for parents and supporting activity planning, so teachers have more time with children. AI should not replace teachers in assessing, classifying or labelling children, and should not be used for emotion recognition or behaviour monitoring of children. Every AI output needs a teacher's review before use.
Disclosure: EYC builds Early Years Circle, an operating platform for early childhood education. This article is product-neutral and applies to any tool.
Where AI helps adults
General principle: AI handles administrative and drafting work; teachers keep decisions and relationships with children and parents. The examples below are illustrative.
- Administrative paperwork: compiling records, filling forms from existing data, reminders.
- Translation for parents: translating notices into the parent's language; the teacher reviews before sending.
- Drafting reports and notices: from a teacher's short notes, AI writes a clear draft; the teacher edits and owns the content.
- Planning support: suggesting activity ideas, material lists and theme adaptations; teachers choose and adjust for the actual children.
Where AI must not replace people
Young children develop quickly and very differently; a wrong judgement can follow a child for a long time. Some work should only be done by qualified adults who know the child.
- Assessing, ranking or labelling children (for example “delayed” or “hyperactive”) from automated output.
- Recognising emotions or tracking children's faces, gaze or attention.
- Independent conversations between children and generative AI chatbots.
- Decisions on a child's safety, health or referral to specialist support.
International guidance: UNESCO and UNICEF
In 2023 UNESCO published the first global guidance on generative AI in education and research. It proposes a human-agency, age-appropriate approach, calls for data privacy protection and proposes an age limit of 13 for the use of AI tools in the classroom, together with teacher training. For preschool children, this means they should not use generative AI tools on their own.
UNICEF's “Guidance on AI and children” version 3.0 (December 2025) draws on the Convention on the Rights of the Child and sets 10 requirements for child-centred AI, including: ensure safety for children; protect children's data and privacy; ensure non-discrimination and fairness; provide transparency, explainability and accountability; and support children's best interests, development and well-being.
Vietnam's 2026 legal framework
AI Law and Decision 33/2026/QD-TTg
The Law on Artificial Intelligence No. 134/2025/QH15 took effect on 1 March 2026. Decision 33/2026/QD-TTg (effective 15 August 2026) issues the list of high-risk AI systems in 6 sectors, including education. In education there are three groups: systems providing self-study content under the education programme using uncontrolled data sources; systems that use AI to automatically test, assess or rank learners; and systems that use AI to monitor and analyse learner behaviour. The Decision's annex sets conditions for each group; according to VOV Giao thông's report, the behaviour-monitoring group applies when biometric data such as face, gaze or emotion is used, or high-frequency automatic alerts cause stress or intrude on privacy.
For listed education systems already operating before 15 August 2026, providers and deployers must complete their compliance obligations before 1 September 2027. The Decision also states that using AI does not transfer or remove the responsibility of the competent person and that human oversight, control and the ability to intervene must be ensured. See Vietnam's AI Law 2025.
Personal Data Protection Law and children's data
The Law on Personal Data Protection No. 91/2025/QH15 took effect on 1 January 2026. Under the law, for children the legal representative exercises the data subject's rights; processing data to publish or disclose private-life information of children aged 7 and over requires the consent of both the child and the representative.
Decree 356/2025/ND-CP guides the law. Article 21(5) lists “collecting and processing personal data via education apps or software with monitoring features such as attendance, video recording, behaviour scoring or emotion recognition” as a personal data processing service; operating such a service requires a certificate of eligibility from the Ministry of Public Security. Schools should ask their software providers about their compliance status. EYC follows these rules closely. See children's data protection in preschools.
Teacher-in-the-loop design
Teacher-in-the-loop means AI only suggests and the teacher decides; anything sent to parents or added to a child's record is reviewed and owned by a person. See human-in-the-loop for the model in detail.
- AI produces drafts and never sends on its own.
- Label which content was AI-assisted.
- Never put children's photos, full names or health information into public AI tools.
- Keep a log of who approved what and when.
Common risks of generative AI in schools
Even when used only by adults, generative AI carries risks that school leaders should explain to teachers. These are not reasons to ban it, but reasons to have usage rules and a review step.
- Wrong or invented content: AI can get health, nutrition or regulatory details wrong; teachers must check against official sources.
- Unsuitable tone: notices to parents need to be warm, clear and match the school's culture; AI drafts are often generic.
- Bias: suggested activities or comments may carry gender, regional or ability stereotypes; they need a pedagogical second read.
- Data leakage: pasting notes containing children's names, health conditions or family circumstances into public tools can expose sensitive data.
- Over-reliance: if AI writes every comment, teachers gradually lose the habit of observing and recording children directly.
Getting started: four steps for school leaders
No big project is needed. A school can start with one small, controlled task and expand once rules and experience are in place.
- Step 1, pick one adult task: for example drafting the weekly notice or translating notices for parents who speak another language.
- Step 2, write short rules: which data must never go into the tool, who approves and how AI-assisted content is labelled.
- Step 3, train teachers: one hands-on session using the school's real examples, stressing review and editing of drafts.
- Step 4, review regularly: ask teachers and parents about quality, time saved and issues, then decide whether to expand.
Checklist for preschools and how EYC approaches this
EYC builds Early Years Circle with four experiences for teachers, families, administrators and classroom TV. We believe AI in preschools should serve adults and keep people at the centre. If your school needs help assessing tools or workflows, see software and SaaS or contact EYC.
| Question | Why it matters |
|---|---|
| Is the tool for adults or used directly by children? | UNESCO proposes an age limit of 13 for classroom AI tools |
| Does it automatically assess, rank or monitor children? | May fall under the high-risk list in Decision 33/2026 |
| Does it include attendance, video or emotion recognition? | Covered by Decree 356/2025, Article 21(5) |
| Where is children's data stored, who can access it and how is it deleted? | Obligations under Law 91/2025 |
| How are parents informed and how do they consent? | Representatives exercise children's data rights |
| Do teachers review every output before use? | Keeps humans in oversight and accountable |
| Have teachers been trained to use it? | UNESCO stresses teacher training |
Frequently asked questions
Should preschool children use AI chatbots?
No. UNESCO proposes an age limit of 13 for using AI tools in the classroom. In preschool, AI should be a tool for adults, while children learn through play and interaction with teachers.
Can AI be used to recognise children's emotions?
Not recommended. AI monitoring learner behaviour with biometric data may be high-risk under Decision 33/2026, and education software with emotion recognition is covered by Article 21(5) of Decree 356/2025.
What should preschool teachers use AI for?
The best fits are cutting paperwork, drafting announcements and reports, translating for parents and suggesting activities. Teachers should read and approve everything AI produces before it is used.
What is the compliance deadline for high-risk AI in education?
For listed systems operating before 15 August 2026, the deadline is 1 September 2027 under Decision 33/2026/QD-TTg.
Can children's photos be put into free AI tools?
No. These are children's personal data, requiring a legal basis and consent under Law 91/2025; public tools make storage hard to control.