Lead
On 16 September 2026, the Chief Executive’s 2026 Policy Address set out AI development and governance in parallel at Policy Address level, announced a Commissioner for AI under the Digital Policy Office, and focused risk governance on seven areas of action. In Chapter 8 of the same document, the government proposed to study and build an “AI City Brain” integrated urban management system—linking works, environment, transport, emergency and related big data, with Phase 1 connecting the Security Bureau’s Emergency Monitoring and Support Centre to strengthen cross-department response and day-to-day city management.
The next day (17 September), Chief Secretary Eric Chan introduced at a press conference that Phase 1 would cost about HK$200 million, target completion in the first quarter of 2028, and stated that the system is not meant to monitor individuals, investigate crime, or “catch thieves”—emphasising better integrated response to emergencies and livelihood matters. news.gov.hk reported the arrangements the same day.
This draft collates announced policy directions, figures and timelines from primary official sources.
1. Policy Address Chapter 3｜Development with governance; Commissioner for AI
Chapter 3 states that the government has pursued AI development and governance in parallel, aiming for technology that is “beneficial” and “safe and controllable” (see §§102, 121–122). Application-related items include finance sandboxes, healthcare coordination, autonomous vehicles, a welfare lab, and AI-for-all training—reflecting a people-centred technology agenda.
§123 adds that the Chief Secretary will coordinate the work, and a Commissioner for AI will be established in the Digital Policy Office (DPO) to assist bureaux in advancing AI application and governance. Grade, recruitment timeline and related details: further details to be announced.
2. Seven risk-governance tracks｜Official order and next steps
Policy Address §122 focuses AI risk governance on seven areas of action—listed here in the official order:
- Combating AI-enabled crime
- Protecting underage users
- Building a governance framework for AI ethics
- Promoting safe application
- Clarifying liability when products cause accidents
- Formulating AI Agent safety management guidelines
- Studying the wider impacts of extensive AI use
Per §123, official arrangements include:
Two easy-to-mix facts:
- LRC deepfake consultation is a planned next step. The Policy Address says views will be sought “in due course.” The LRC’s January 2026 report on cyber-dependent crimes (Part 1) is a separate evidence chain—do not conflate it with the deepfake “cyber-enabled / timely consultation” track.
- The seven-track list appears in the Policy Address; items are progressing through study, pilot, consultation or forthcoming-publication stages.
3. “AI City Brain”｜Stronger urban governance and emergency response
What the Policy Address says (Chapter 8)
Chapter 8 §§418–420 proposes to study and build an “AI City Brain”: linking and integrating big data from works, environment, transport, emergency and related urban-governance departments; a task force chaired by the Chief Secretary, with Security, ITIB, Development, and Transport and Logistics among members.
Phase 1: the Security Bureau Emergency Monitoring and Support Centre’s digital infrastructure will connect to the system; AI will push risk alerts and trend forecasts to strengthen cross-department responses to extreme weather, major incidents, and day-to-day public-order management, later advancing toward comprehensive urban management. Related measures include enhancing the Common Spatial Data Infrastructure / 3D digital map / digital twin; a Police “inter-departmental major incident platform”; a Transport Department traffic management platform under study (Aberdeen pilot—extreme-weather functions in 2027, remaining functions in 2028); and expansion of Development Bureau’s SmartAERO drone joint-control platform to strategic areas to help handle landslips, flooding, fallen trees and similar emergencies faster.
The same chapter also includes related improvements—e.g. §434 on AI detecting repeat 1823 complaints, moving from post-complaint handling toward proactive prevention; and §439 on expanding drones for emergency crime-prevention to better protect the public. news.gov.hk summarised City Brain and related measures on 16 September (TC piece).
Official timeline and budget (17 September)
Per news.gov.hk, 17 September 2026:
- Task force targeted for Q4 2026;
- Phase 1 base = Government Cloud; scenario = Security Bureau EMS Centre;
- Priority connections include CEDD landslip warning and DSD flood monitoring systems, among others;
- Phase 1 cost about HK$200 million; target completion Q1 2028;
- Q3 2028 priority hook-up of 1823 and emergency-related data, so the public can see practical results sooner;
- Full maturity about six to nine years (drawing on Shanghai, Hangzhou and similar experience);
- Government will ensure Phase 1 is delivered and bedded in before stepwise expansion and orderly connection of departmental systems.
Stated purpose safeguard
Chief Secretary Eric Chan stressed that the system is not for monitoring individuals, investigating crime, or “catching thieves”; the policy focus is interconnecting departmental data for faster, more accurate monitoring and analysis, early risk identification and alerts, and stronger integrated response to emergencies and livelihood matters. That statement of policy intent at the press conference (as reported by news.gov.hk) is the government’s positive purpose safeguard for the system. Written task-force terms, data arrangements and similar detail: further details to be announced.
Do not mix the money: City Brain Phase 1 ≈ HK$200 million ≠ Welfare AI Lab HK$300 million (see next section).
4. Welfare AI Lab and AI for All｜People-centred complements
Welfare AI Lab: HK$300 million
§113: the government will allocate HK$300 million to set up WelfareHK.ai, identify application scenarios, and fund development and trials—including analysing service use at welfare facilities to match needs more precisely. Operational trial details: further details to be announced.
AI for All: 200+ courses/activities—and a separate “at least 40,000 employees” plan
§118: under “AI for All,” more than 200 training courses and activities by Q1 2028. §119 adds a “robots in the community” trial; §120 an 18-month AI skills plan under the upgraded skills body, expected to benefit at least 40,000 employees—a figure that must not be collapsed into the “200+ activities” metric.
The LWB employment-impact report is “to be published soon” (§123).
5. Announced next milestones
Against the published timeline, items to watch include:
- Establishment of the Commissioner for AI and related mandate (further details to be announced)
- AI City Brain System Task Force targeted for Q4 2026
- LRC public consultation on deepfake and related issues “in due course”
- City Brain Phase 1 target completion Q1 2028; Q3 2028 priority connection of 1823 and emergency data
- Welfare Lab and AI-for-All training progressing to the targets set out in the Policy Address
