How Hong Kong’s government is using AI

The confirmed picture

Where AI is being used now

1. Internal productivity: statistics, extraction and analysis

The 2025 Policy Address frames AI as a way to improve the efficiency of government work. It gives the Census and Statistics Department as a concrete example: the department has applied AI to examine trade data and retrieve financial information from the financial statements and announcements of thousands of listed companies for statistical compilation.

That is a classic public-sector AI use case: large volumes of information, repetitive extraction tasks and relatively structured documents. It is not evidence that professional statistical judgement has been replaced across the board. It does show AI being inserted into defined, data-heavy processes. Data analysis is also one of the priority AI application areas named for wider government use.

2. Public service: 1823, speech enquiries and drafted replies

The clearest public-facing example is 1823, the government service used for public enquiries. Official material says 1823 uses AI to automatically identify citizens’ voice enquiries and to assist in drafting written replies. The government says the related processing time has fallen by about 30%.

The cautious reading is important. The public record supports AI being used for tasks such as enquiry handling, speech recognition, drafting support and repetitive customer-service workflows. It does not prove that AI is directly deciding every citizen case or replacing all human judgement in 1823 or other public services.

3. Document processing: a stated next priority

Beyond statistics and customer service, document processing is explicitly listed as a priority area. The 2025 Policy Address says the Digital Policy Office will roll out AI applications in data analysis, customer service and document processing, promote wider use across departments, and keep the applications secure and controllable.

That means the government’s AI agenda is not limited to chatbots or telephone support. It also reaches back-office administrative work. However, the public documents reviewed here do not provide a full list of every department, system, document workflow or implementation timetable.

Smart Government Innovation Lab: test first, then scale

Hong Kong’s approach appears to be incremental rather than a single big-bang replacement of government systems. Legislative Council material says the Digital Policy Office is advancing digital government and smart-city work by strengthening digital infrastructure, promoting data flow and exchange, and driving innovation and smart applications inside government.

The Smart Government Innovation Lab is one mechanism for this. Public material says that, over the past three years, it matched more than 70 solutions for over 20 government departments and completed more than 40 proof-of-concept projects. Named examples include an AI chatbot for the Working Family and Student Financial Assistance Agency and video analytics technology for the Social Welfare Department.

That points to a trial-and-evaluate model: identify a departmental problem, run a proof of concept, and then consider whether the solution can be built into regular operations.

Central oversight: from scattered pilots to a government-wide push

The 2025 Policy Address announced the AI Efficacy Enhancement Team to co-ordinate and steer departments in applying AI effectively. A separate Policy Address section says the government attaches great importance to AI governance and risk management, and that, under the team’s leadership, AI applications will be promoted in government services.

This matters because it moves AI adoption beyond isolated departmental experiments. At the same time, the official language keeps returning to safety, controllability and risk management, rather than treating AI as a technology that can simply be dropped into any process without controls.

Guardrails: not every dataset can go into any AI tool

Hong Kong’s government is not only talking about adoption. It is also defining boundaries. On April 15, 2025, the Digital Policy Office released the Hong Kong Generative Artificial Intelligence Technical and Application Guideline. The government said the office had earlier commissioned the Hong Kong Generative AI Research and Development Center to study and suggest codes and guidelines covering accuracy, responsibility and information security in generative AI technologies and applications.

The guideline is aimed at technology developers, service providers and users, and is positioned as practical operational guidance for applying generative AI technology. It also says organisations adopting generative AI services should develop internal policies or guidelines covering permitted tools, permitted uses, policy applicability, the permissible types and amounts of input information, how output information may be used, and how output information may be stored.

In plain English, the direction is not: paste anything into any public AI system. It is: define the approved tools, approved purposes, input limits, output use and storage rules before adoption.

There is also a broader governance layer. The Digital Policy Office says the government formulated the Ethical AI Framework in 2021 to provide guidance for projects involving AI technology. The framework covers areas such as AI governance structure, the AI lifecycle, practical guidance, AI assessment and AI application impact assessment, giving project teams a reference from planning through operation and monitoring.

Training: civil servants need AI capability too

AI deployment is not just a procurement exercise. It requires people who understand technology, data management, cybersecurity, project governance and accountability. Legislative Council material says the Civil Service College will continue strengthening technology-application elements in leadership training, helping departmental leaders understand how technological development affects policy formulation and implementation, and how to use big data and AI for public service innovation.

Another Legislative Council paper says the Digital Policy Office and the Civil Service College regularly organise thematic seminars under the Innovation and Technology leadership series for senior management across bureaux and departments, covering project governance, cybersecurity and AI. The Digital Policy Office also says that since late 2024 it has worked with the Civil Service College on briefings, seminars and training to strengthen senior government personnel’s core skills in information technology, data management and cybersecurity.

What should not be over-read

First, the public record should not be read as proof that AI is about to replace all civil-service decision-making. The clearly documented areas are data analysis, customer service, document processing, statistical data extraction and selected departmental smart-application pilots.

Second, the generative AI guideline is not a blanket approval for staff or organisations to use any AI tool in any way. It requires organisations to specify permitted tools, permitted purposes, input information limits, output use and output storage arrangements.

Third, the public material reviewed here does not amount to a complete master list of all AI deployments across every department, with full budgets, timelines and performance indicators. If AI is later used in more sensitive areas such as approvals, enforcement or policy-decision support, those deployments would need to be checked against departmental papers, procurement documents, Legislative Council records and published performance measures.

Bottom line

The most accurate answer is that Hong Kong’s government is using, and planning to expand, AI in four main directions: internal administrative efficiency, public customer service and digital services, data analysis, and document processing. It is pairing that rollout with the AI Efficacy Enhancement Team, the generative AI technical and application guideline, the Ethical AI Framework and civil-service training.

What remains unclear is how deeply each department will eventually embed AI, whether the technology will move into higher-risk decision-support settings, and how future budgets, procurement details, KPIs and accountability arrangements will be disclosed.