Following is a question by the Hon Tommy Chung and a written reply by the Secretary for Innovation, Technology and Industry, Professor Sun Dong, in the Legislative Council today (July 15):

Question: 

     In recent years, the Government has been actively promoting the development and application of AI technology, covering areas such as enhancing public services and building an industry ecosystem. The Digital Policy Office earlier published application guides such as the Ethical Artificial Intelligence Framework and the Hong Kong Generative Artificial Intelligence Technical and Application Guideline, and developed relevant guidelines in respect of information security, thereby providing a governance foundation for the development and application of AI technology. There are views that providing robust infrastructure for AI technology and regulating its use are key factors in facilitating the healthy and long-term development of AI technology. In this connection, will the Government inform this Council:

(1) whether it has compiled statistics on the growth trend of the local daily average token consumption in the past two years; whether it has projected the estimated growth rate of and demand for the daily average token consumption in the next three years, as well as the token capacity of local AI infrastructure; if not, the reasons for that and whether the relevant statistics will be compiled;

(2) whether it has formulated specific operating standards for government departments and public organisations regarding the application of AI technology, including account registration procedures, capability assessments and testing of AI application tools, a whitelist of AI application tools, as well as content moderation and accountability mechanisms; if so, of the details; if not, the reasons for that; and

(3) whether the Government has internally formulated vetting criteria and accountability mechanisms in respect of promotional videos generated using AI tools, including verification of content authenticity, assurance of non-infringement of copyright, and screening standards relating to sensitive information and values; if so, of the details; if not, the reasons for that?

Reply:

President,

     AI is a core driving force leading the new round of technological revolution and industrial transformation, as well as an important engine pushing forward the development of the digital economy and smart city. In accordance with the Hong Kong Innovation and Technology Development Blueprint, the Government is actively building a local AI ecosystem and provides clear guidelines for various sectors of the community to develop and apply AI technologies through the Digital Policy Office (DPO). Reply to the three parts of the question is as follows:
 
(1) A token is the basic unit used by AI systems to process text, and is commonly used to reflect data processing volume and the related demand for computing power. Currently, government departments, scientific research institutions and enterprises directly procure services from local, Mainland or overseas suppliers based on their business needs. Their token consumption data is scattered across different platforms and proprietary networks. As a unified standard for the calculation method of tokens when different models process the same text has yet to be established in the global AI industry, the Government currently has no statistics or projections on the local daily average token consumption and capacity.

     To seize the opportunities brought by the rapid development of AI, the Government has been actively deploying relevant digital infrastructures. Hong Kong's overall computing power has currently reached 5 000 peta-floating point operations per second (PFLOPS), with Cyberport's AI Supercomputing Centre accounting for 60 per cent of it. In addition, the Sandy Ridge Data Facility Cluster, which has entered the construction phase, is expected to provide 180 000 PFLOPS of computing power by 2032, equivalent to 36 times Hong Kong's current computing power, providing robust support for AI computing demands including token processing.

     The Government will closely monitor the latest global and local AI development, and continuously optimise various digital infrastructures.

(2) and (3) While promoting the innovative application of AI, it is essential to effectively manage technical risks and ensure that operations comply with safety and ethical standards. The DPO has published several key guidelines, including the Ethical Artificial Intelligence Framework, and the Hong Kong Generative Artificial Intelligence Technical and Application Guideline targeting generative AI technologies. These guidelines are applicable within the Government and also serve as a reference for adoption by public organisations as well as various sectors of the community. The Office of the Privacy Commissioner for Personal Data has also published relevant guidelines, such as Guidance on the Ethical Development and Use of Artificial Intelligence, Artificial Intelligence: Model Personal Data Protection Framework and Checklist on Guidelines for the Use of Generative AI by Employees, to provide practical recommendations on AI governance and personal data privacy protection.

     The Government has formulated relevant regulations regarding the application of AI, the protection of personal data privacy, and information security for departments. In addition to the overarching guidelines mentioned above, the DPO has also published the Artificial Intelligence Adoption Guide for Government Bureaux and Departments in February 2026. The guide stipulates that content generated by AI tools (including videos and images) must be verified for its authenticity through official or trusted sources; and prohibits the input of confidential or personal data into external or public AI platforms and instructing the generation of content that infringes copyright or intellectual property rights. The Government Information Technology Security Policy and Guidelines also establish standards for account registration and management, as well as system operations, including the encryption of data in transit and at rest, and require various departments to regularly conduct security risk assessments and audits on their information systems and relevant data.

     To encourage departments to explore AI applications under safe and controllable circumstances, promote workflow optimisation, and save the time and administrative costs for departments to procure separate systems, the DPO has introduced the "AI Solution Catalogue" and "AI Toolbox". These consolidate over 260 AI solutions that have undergone technical screening and are applicable to government scenarios (such as text summarisation, translation, data analysis, and official document drafting support), with details on their use cases, technical architecture and limitations.

     Furthermore, the Government continues to strengthen talent training to enhance the digital mindset and AI application capabilities of civil servants. In collaboration with the Civil Service College, the DPO has launched tiered online AI training courses covering areas such as foundational knowledge, ethical safety, tools application, as well as frontier technology trends and regional developments, accompanied by assessments to evaluate and reinforce learning outcomes. Regular thematic seminars under the "Innovation and Technology leadership" series are also organised to strengthen digital governance capabilities at the management level. In 2025-26, the DPO has arranged around 130 training programmes related to digital technologies, promoting the integration of theory and practice and nurturing a civil service adept in the use of technologies.

     Both in guidelines and in training, the Government emphasises the importance of a human-in-the-loop approach. The relevant guidelines clearly state that AI tools serve only an assistive role, and that organisations adopting such technologies bear the power and responsibility for the governance and verification of their outputs, and must ensure compliance with relevant ethical principles. Accordingly, before publishing any AI-assisted content, government departments should conduct human review and verification to ensure its accuracy and authenticity, and not rely solely on AI-generated outputs.

Ends/Wednesday, July 15, 2026
Issued at HKT 14:45

NNNN