On 17 July, at the opening of the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance in Shanghai, President Xi Jinping delivered the keynote "Working Together to Build a Fair and Reasonable Global AI Governance System", requiring that AI — the "thousand-li horse" — runs both fast and steady, that AI remains a trustworthy tool always under human control, and opposing the broad application of national-security concepts in the AI domain.
This launch was not a hasty response but the inevitable outcome of long-term strategic study by the "GBA AI-Safety Industry–Academia–Research Coordination Mechanism". Half a year earlier the mechanism had already run deep scenario exercises on cognitive-infiltration risks arising from AI "survival motives", and in April this year it submitted a dedicated risk-assessment report to national security authorities. Only recently, as the core theory reached top preprint platforms and technical maturity hit the industrialisation inflection point, did the mechanism choose this moment to launch jointly with industry, academia and research partners.
The President's speech set out four positions — open innovation, safe and controllable, mutual learning among civilisations, and global governance — with "strengthened risk and safe control" pointing directly at education, AI's most sensitive deployment scenario. The Initiative responds to the Ministry of Education's "AI + Education" Action Plan requirement to "build a safety-protection system for AI in education", prioritising three lines of defence while AI empowers K12 research and teaching: child data safety, content compliance, and values guidance.
The Initiative's technical foundation stems from the HoloTSH (Holographic Tensor-Spectrum Hypergraph) school founded by core mechanism members, led by Prof. Ng Kwok-Po — Chair Professor and Dean of Science at Hong Kong Baptist University, SIAM Fellow and Feng Kang Award laureate. Building on his foundational work in tensor singular value decomposition (t-SVD), the team constructed HoloBlocker (topological circuit breaker) and HoloAuditor (holographic audit architecture), pioneering Block Tensor Decomposition (BTD) to solve algorithm black-box and heterogeneity problems at the level of mathematical axioms — turning "safe and controllable" into verifiable, auditable standards.
The President's speech announced the founding in Shanghai of the World Artificial Intelligence Cooperation Organization (WAICO), led by China. Dr. Ho Kin-Chung, President of HKAAST, said that the combination of "topological circuit-breaking + holographic audit" builds a national-security outpost on the mathematical foundation, and by aligning with standards of international bodies such as the Linux Foundation, exports a replicable "GBA experience" for national AI governance.
Founded in 1985, the Hong Kong Association for the Advancement of Science and Technology is a non-profit professional body of Hong Kong's science community. Under the philosophy of "vertical and horizontal linkage, cross-border synergy", it helps Hong Kong integrate into the nation's AI development and safeguards national data sovereignty and ideological security.
Media enquiries: HKAAST Secretariat