A student came back to see me over the holidays. He is at university in the United States now, reading Economics, and he comes to see me whenever he is home.
He mentioned that one of his courses is on AI programming. I asked how they learn it.
They start with some basic programming, Python and so on, but only far enough for him to understand the structure and to read what the AI produces. Then very quickly they move onto real cases, building things with the AI helping. What they are actually practising is how to communicate with the AI so that it produces the result they want.
He is reading Economics, not computer science.
After he left, a question stayed with me. Is anything like that happening in Hong Kong schools? Is there any government guidance? So I went and read the documents.
What stayed with me afterwards was not the funding or the administration. It was what the documents say about how students are supposed to learn.
What the Blueprint requires schools to do
On 17 June 2026 the Education Bureau issued Circular No. 11/2026, announcing the Blueprint for Digital Education Development in Primary and Secondary Schools, compiled by the Curriculum Development Council. On 25 August it issued Circular Memorandum No. 156/2026, setting out the arrangements for the 2026/27 school year.
Between them, what schools have to do this year is clear:
- Put it in the plan. Every primary and secondary school is to incorporate digital and AI education strategies into its School Development Plan, set a clear school-based implementation timeline, and drive a student-centred shift in learning, teaching and assessment.
- Set up a task force. A school-based digital education task force to lead and coordinate AI education within the school.
- In the curriculum. AI learning elements are to be integrated into Key Learning Areas, subjects and cross-disciplinary domains — "AI + Curriculum".
- Train teachers. At least 30 hours of digital education training within each three-year CPD cycle of 150 hours. All teachers take basic "AI Literacy" and "AI + Subject" training; principals and coordinating teachers also take "AI Leadership". No fewer than 50,000 training places a year.
- The funding. About $500 million under the Quality Education Fund, giving every publicly funded school a one-off $500,000 grant, usable across 2025/26 to 2027/28 for AI teaching tools and resources and for subsidising students' participation in related activities. How it is spent goes into the School Report, is approved by the management committee, and is uploaded to the school website.
The biggest change this year is that none of this is merely encouragement any more. It goes into the School Development Plan, and it connects to the school development and accountability review.
One thing is worth clearing up, because this is where most parents have it wrong.
At secondary level, AI is not a new subject and there is no AI exam. It is folded into the existing curriculum. The one genuinely new standalone framework — the "Preliminary Study of Artificial Intelligence" strand of the Information and Innovation Technology Curriculum Framework — is for primary schools, and is a pilot.
What the documents say about how students should learn
This is the part that stayed with me.
Among the Blueprint's overall goals is an explicit commitment to the human mind as the mainstay with the computer in a supporting role, on a foundation of cultivating values and nurturing people. The learning approach it describes is captured in one phrase — "learn to apply, apply to learn, learn to think, think to innovate" — learning through practice and deepening understanding through application.
The August circular is more direct still: schools are to teach students the limitations of AI, students should not rely on it blindly, and they should not hand over important parts of learning — collecting information, analysing it, reflecting on it — to AI tools, while strictly upholding academic integrity. Deepfakes and data bias are flagged too.
In other words, the official position is not that students should use AI less. It is that they should not hand over the thinking.
What my student is learning in the United States is the other side of the same thing. He is not practising how to write complicated code. He is practising how to direct a tool, and how to read what it hands back.
The two most useful things in the documents, for a parent
First, a word about what the documents don't say. The same phrase runs through all of it: school-based implementation strategies, school-based timelines, according to school context. Two schools can both comply fully and produce completely different things. So the question of what your child will learn can only be answered by your school.
With that in mind, two things in the documents are genuinely useful to a parent.
First, schools are asked to run parent education. The documents talk about deepening home-school co-operation and organising parent education activities so that parents can guide their children to use AI responsibly. You have grounds to ask whether your school has anything planned this year.
Second, the school's plan and its spending can be checked. Digital education goes into the School Development Plan or Annual School Plan, and the interim and final reports on the $500,000 go into the School Report and onto the school website. If you want to know what your school is doing, that is the most concrete way in.
One more thing, because it affects a fair number of families. The August circular memorandum is addressed to school heads "excluding ESF Schools and International Schools". If your child is at an ESF or international school, these arrangements do not apply directly, and you should ask the school.
What I think
To be clear first: this policy is very new and a lot of the detail is still unclear. What follows is what I have picked up from my own students, and it cannot stand for every school in Hong Kong.
A lot of the secondary students I see start from the very bottom, whether the subject is called AI or ICT — Python or another language, or basic computing knowledge.
I am not saying the basics don't matter. They do.
Mathematics is, in large part, a training in thinking. And it is built in layers. If the algebra is weak, what comes after it is hard to follow. If the lower layer is unstable, there is nothing to build on.
AI is not like that. AI rewards practice and application, the content changes constantly, there is no point at which you have learned enough, and not knowing one part does not stop you learning another. It does not stack the way mathematics stacks.
But the two share something, and I think that is where the point actually is. We want a student to genuinely understand what they are learning, not merely to be able to say it, recite it, and reproduce it in an exam.
So the aim of studying AI or ICT should not be to write "what AI is" or "why AI matters" on an exam paper. It should be to use AI to actually get something done, and to make something.
Knowing what you want to make, and learning how to talk to the AI so that it gets made — those are the two that matter.
That Economics student is learning exactly that.
On direction, the Blueprint and I are not far apart. The difference is only in what that turns into inside each individual classroom.
One thing to try this week
Go to your school's website and look for the School Development Plan or Annual School Plan, and find the section on digital education.
If it's there, you'll know what the school intends to do this year. If it isn't up yet, that's a good question for the next parents' day.