Founded and led by HKAAST · Secretariat provided by HKAAST Greater Bay Area · 3 Locks: Safe · Reliable · Trustworthy 繁 · 简 · EN Want to know how the National AI Safety Governance Framework 3.0 lands in your industry?
GBA AI Safety & Development Standing CommitteeNon-profit AI Safety Think Tank

Services & Industry–Academia Partnership

The Committee is a non-profit think tank. The four cooperation tracks below all grow out of our research — assessments run on our own measurement engines, governance on our published architectures, and income supports public-interest research and education.
Our stance: we sell measurement, not shields.

We never promise "absolute safety" and we never sell fear. What we provide is evidence and architecture that is measurable, auditable and fresh — the decision to allow, encrypt, localise or keep data in-border always stays with the institution. Scenarios we cannot measure are honestly labelled INCONCLUSIVE.

Four cooperation tracks

01 · Audit

AI-Safety Assessment

Empirical measurement on the SA-RI audit engine, providing evidence for data export and model deployment:

  • Representation-leakage audit for split learning / split inference (dual attack families, cross-metric)
  • Data-export security reports: AMBER/RED grading by fixed engine thresholds, not client-chosen
  • Four report classes: training-track / inference-track / export audit / edge deployment
  • Results with variance and boundary conditions; unmeasurable = INCONCLUSIVE
02 · Governance

Governance Deployment (Governor)

Turning research architectures into operational governance gates:

  • Agent governance family: HoloManager behaviour steering (steering wheel) + HoloGovernor side-channel monitoring (speed limiter + dash cam)
  • AI behavioural-boundary design: least-privilege "badge", full-process audit trail
  • Topological circuit breaker (HoloBlocker): millisecond fusing on boundary breach
  • Content-safety gate (HoloAegis) with alert-mode risk warnings
  • Compliance alignment for education, healthcare and industrial data spaces
03 · Research

Commissioned Research & Joint Labs

Industry–academia collaboration with our university and institute partners:

  • Commissioned studies on AI safety, data governance and trustworthy intelligence
  • Joint labs and shared research programmes (HoloTSH school directions)
  • Trusted data space co-builds: onboarding review, controlled tasks, metering & reporting
  • Joint publication: preprints, whitepapers, and open-source releases
04 · Education

Training & the School Network

Our non-profit education and talent mission:

  • GBA school K12 research network and the ShenNong AI ecosystem
  • AI-safety workshops for teachers and institutions (literacy course modules)
  • Technical support for implementing Hong Kong's Digital Education Blueprint
  • Free public resources: literacy articles and the AI Safety Assistant

How cooperation works

1 · Contact the SecretariatEmail info@3token.org with your needs and scenario
2 · Needs analysisWe assess the measurable scope and give honest feasibility boundaries
3 · Cooperation planAgree report classes / governance scope / research topics and division of work
4 · ExecuteAfter signing, we deliver evidence and architecture, with re-measurement as needed
🤝 Partners: the Committee works with the University of Macau, Hong Kong Baptist University, The Education University of Hong Kong, Southern University of Science and Technology, Guangdong University of Technology, the Guangdong Digital Industry Research Institute and more — with Huawei and Linux Foundation APAC as advisors and 19 corporate members. See Committee structure.