A milestone moment for Hong Kong's digital education.
On 17 June, the Hong Kong SAR Education Bureau formally released the Development Blueprint for Digital Education in Primary and Secondary Schools. Prepared by the Curriculum Development Council in response to the Chief Executive's 2025 Policy Address, the Blueprint sketches the path to full digital and AI transformation of basic education through "four priorities · ten strategies".
The Blueprint marks Hong Kong's move from "tool adoption" into the deep water of "literacy reshaping". Yet opportunity comes with challenge: as AI enters the classroom at unprecedented depth, student-data safety, algorithmic bias and the digital divide all sharpen. How do we build a solid, trustworthy "Digital Great Wall" for the next generation while embracing the AI dividend?
This is where the GBA's AI-safety research community can shine. The Guangdong Digital Industry Research Institute (GDIRI), with university teams including The Education University of Hong Kong and Hong Kong Baptist University, is combining frontier "topological safety" theory with AI-literacy education — technical support and a working paradigm for implementing the Blueprint.
At its core the Blueprint is four words: literacy, safety, equity, effectiveness.
It calls for AI literacy to enter the core framework of teacher and student digital capability, with deep integration of AI into primary and secondary curricula. Students must not only use AI but learn to handle it — understanding its principles, recognising its limits, guarding against its risks.
This shift chimes with the team's long-advocated "trustworthy AI education". As the team lead puts it: "The second half of digital education is not a contest of who uses AI fastest, but who uses it most steadily and safely."
As digitalisation deepens, campuses become data-intensive — student profiles, health records, learning-behaviour trails. Traditional defences (firewalls, encryption) target external attack but struggle with internal leakage and compliance blind spots. The GDIRI-led team brings "topological data analysis" into education-data governance:
1. Monitoring the "shape" of data flows: topological modelling of campus data streams detects anomalous "holes" or "distortions" in data structure, exposing covert leakage and misuse.
2. Minimisation by design: borrowing the "tensor–hypergraph" audit framework proven in medical data, every collection and call of education data carries tamper-proof mathematical vouchers — "data usable yet invisible".
3. Landing privacy-enhancing technologies (PETs): federated learning, differential privacy and peers, packaged as standardised toolkits schools can "afford and understand".
Pilots in GBA schools are showing early results — a complete loop from concept to practice for the Blueprint's data-safety promise.
Technology is cold; education is warm. The Blueprint's deeper meaning is empowerment as much as protection — raising a generation that commands AI with respect.
With EdUHK colleagues, GDIRI is developing AI-safety literacy course modules for primary and secondary schools:
1. Scenario-based teaching: immersive cases such as "spotting AI face-swap scams" and "runaway-agent simulations" let students feel risk in realistic settings.
2. Open-source tool practice: students run attack-and-defend drills with open security tools, turning "safety as habit" into muscle memory.
3. Ethics from the start: youth versions of frontier debates — "AI survival motives", "algorithmic bias" — spark early thinking about tech ethics.
The Blueprint paints a grand vision; delivering it needs government, schools, researchers and industry in resonance.
With the "shield" of topological safety and the "spear" of AI literacy, the team stands ready to build the Digital Great Wall with Hong Kong's educators — so every child can explore the digital era safely and grow with confidence.
Technology for good; wisdom lights the future. Not a slogan — our shared action plan.
The Guangdong Digital Industry Research Institute is a leading GBA institution in the digital economy, with breakthroughs in cloud disaster recovery, "East-Data-West-Compute" implementation and invisible data-exchange platforms, and is committed to smart-city development and AI talent cultivation (the "Digital Shaolin" programme).
Hong Kong's Digital Education Blueprint sets out "four priorities · ten strategies"; this commentary pulls attention back to an angle often missed: students should not just be AI users, but risk-aware handlers. The gap between "learning AI" and "using AI safely" is exactly what the Committee's public education fills — this site's AI Safety Assistant and beginner articles are open materials built for that purpose.
The phrase "data usable yet invisible" deserves every school leader's attention: campuses are among the most sensitive personal-data environments (student profiles, health records), and firewalls aimed at outsiders do not cover internal compliance blind spots. Mathematical audit that leaves traces on every data flow is a direction schools can adopt step by step, even on tight budgets. Further reading: our five AI pitfalls, data-export primer, and the K12 research-network preprint (ChinaXiv 202606.00271) on our publications page.