Hong Kong's new school year began on 1 September 2026, with artificial intelligence education moving from a policy document into school-based implementation. For parents, the central question is not whether a child can write a more polished prompt. It is whether faster answers still leave the child able to explain the purpose of a question, identify what must be checked, protect information that should not be shared and take responsibility for the final choice.
In practical terms, children need five connected capabilities in the age of AI: framing a problem, understanding the tool, verifying sources, making responsible choices, and testing an answer in the real world before reflecting on what happened. AI can shorten the search for an answer. It cannot decide which problem is worth solving or what fits the situation in front of the child.
What changes for Hong Kong AI education in 2026/27?
The Education Bureau announced its Blueprint for Digital Education Development in Primary and Secondary Schools on 17 June 2026. On 25 August, its Blueprint page listed arrangements for school-based planning and teacher training in the 2026/27 school year. On the first day of school, the Secretary for Education again referred to the Blueprint, digital transformation and related funding.
The Blueprint's Executive Summary says all primary and secondary schools should include digital and AI education strategies in their development plans. It calls for the AI Literacy Learning Framework to be implemented across learning inside and beyond the classroom, with attention to critical thinking, creativity, problem-solving, self-directed learning, cyber security and privacy.
It also sets out a principle that deserves particular attention: the human mind remains the mainstay, with the computer in a supporting role.
This direction is not simply a plan to add more devices, nor does it require every lesson to use generative AI. A policy objective, a school's implementation and a student's eventual learning are three different things. Schools will make choices according to their learners, curricula and readiness. Parents need to understand the rules in their own child's school rather than read “AI for all students” as permission to hand every answer to a system.
Using AI is not the same as possessing AI literacy
AI literacy is broader than operating one chatbot. Tools change, interfaces are redesigned and today's familiar feature may soon be replaced. If learning stops at buttons and prompts, the capability may expire with the product.
The UNESCO AI Competency Framework for Students organises twelve competencies across four dimensions: a human-centred mindset, ethics of AI, AI techniques and applications, and AI system design. The progression runs from understanding to applying and creating. It emphasises critical judgement, human accountability and students as responsible users and co-creators.
On 20 August 2026, new US Department of Education guidance on classroom technology offered another practical test. Before selecting a tool, ask what learning problem it solves, when it should be used, for whom, for how long and what evidence shows that it improves learning.
These frameworks are not identical. They do share one boundary: technology is a tool; the purpose of learning and human judgement should not disappear behind the screen.
Five capabilities children still need
1. Frame the problem before opening the tool
A complete prompt does not necessarily show that a child understands the problem. The first task is to state: What am I trying to find out? Why does it matter? How will the answer be used? Which conditions cannot be ignored?
If a child is planning a museum visit, “make me an itinerary” is not the end of the task. They still need to consider who is going, the time and budget available, whose needs must be included and which details require confirmation on official pages.
2. Keep enough knowledge to judge the answer
AI can organise, rewrite and generate options, but fluency is not accuracy. Children need foundational knowledge to notice that a number is implausible, a causal claim moves too quickly or an answer has missed the Hong Kong context.
AI therefore does not make reading, numeracy, historical context or scientific concepts less valuable. When answers become easier to generate, the knowledge required to judge them becomes more important.
3. Verify sources and protect privacy
An AI-generated sentence is not a source. Dates, rules, opening hours, transport, health and safety information should be checked against the responsible authority, an original document or another credible source. Children also need to distinguish a verifiable fact from an inference or recommendation.
A child's full name, school, photograph, home address, live location, schedule, health information or account details should not be entered into an unapproved tool. Younger children need adult support. For schoolwork, the school's rules on AI use, attribution and original work come first.
4. Choose, and notice who the choice affects
Real problems rarely have one universal “best” answer. An arrangement may be fastest but unsuitable for a younger sibling. A recommendation may be cheapest only because it transfers the burden to someone else. AI can list options; a family still has to decide what matters, which trade-offs are acceptable and who will carry the consequences.
Agency does not mean rejecting all support. It means that, after receiving information, guidance and adult boundaries, a child can still explain why this is the next step they have chosen.
5. Bring the answer into the real world, then reflect
A plan encounters time, weather, other people's needs, physical experience and unexpected change only when it is used. Children need to discover that an unsuitable answer can be questioned. They can say that they do not know, reformulate the problem, change method or ask another person for help.
This is where sisu meets AI literacy. Sisu is not a refusal to use a new tool, nor an instruction to persist with a wrong answer. It is the ability to retain judgement and find the next appropriate action when the situation becomes uncertain. Reflection after action helps turn an episode of tool use into real-world learning.
A family AI judgement card
The next time a child wants to use AI for schoolwork, research or a small family plan, work through these five questions. This is a reflection tool developed for this article, not an official Education Bureau or UNESCO assessment.
- Purpose: what problem are we trying to solve? If AI were unavailable, where would we begin?
- Evidence: which parts of the answer require verification? Where is the closest primary or official source?
- Boundaries: what does the system not know, and what information should we not give it? Do age, safety, privacy or school rules require an adult decision?
- Decision: which part can the tool support, and which part must a person choose and own? Who else will be affected?
- Reflection: what happened after we used the answer? What helped, what was inaccurate and what should change next time?
The card need not become a test after every use. A parent might ask only, “Which part of this answer most needs checking?” A child might compare two sources, or state an initial view before asking AI to produce a counterargument.
When a task involves an outdoor route, health, a hazardous environment or serious consequences, AI can help organise questions but cannot replace qualified advice, current official information or an adult's safety responsibility. Weather, equipment, venue, supervision, withdrawal conditions and emergency arrangements still require confirmation by people who can carry that responsibility.
A parent's role is not to find every answer
A total ban on AI may remove the chance to learn its use with guidance. Giving every assignment and decision to AI may produce a finished-looking answer without understanding. Both extremes avoid the harder work: establishing boundaries together and gradually returning more judgement to the child.
An appropriate challenge does not leave a child without support. Adults remain responsible for risks, privacy and consequences beyond the child's capacity. Within clear boundaries, the child can practise asking, checking, choosing and revising.
Support need not take the form of another answer. A parent can ask:
- What do you already know, and what is still missing?
- Is this the system's statement or a verifiable fact?
- Which part will you use, and why?
- If reality turns out differently, how will you adjust?
Three short answers about AI education in Hong Kong
Are Hong Kong schools using AI across the board from 2026/27?
Hong Kong has entered a system-wide implementation phase for AI education. Primary and secondary schools are expected to include digital and AI education strategies in their development plans, supported by student frameworks, teacher training and resources. This does not mean that every school, subject or assignment will use generative AI in the same way.
Is AI literacy the same as prompt writing or coding?
No. Prompt writing and coding may be components. A fuller definition includes understanding basic principles and limitations, verifying output, protecting privacy, recognising bias, considering ethics and impact, and preserving human responsibility and choice.
What does sisu have to do with AI education?
AI can provide options quickly. Sisu becomes relevant when those options are incomplete, the context changes or a first attempt fails. It helps a person continue to observe, judge, seek support and adjust. It does not compete with technology; it keeps the human being in the active position.
Faster answers require space for slower judgement
The new school year's direction is worth welcoming and implementing with care. Children do need to learn to use contemporary tools. Education, however, cannot be measured only by the number of AI interactions or by how complete an output appears.
What matters more is whether a child can name the problem, locate the evidence, protect themselves and others, make a choice and learn again when reality differs from the forecast.
AI can bring an answer to the child.
What cannot be outsourced is how they understand it, which choice they are prepared to own and how they intend to take the next step.