Hong Kong's Secretary for Innovation, Technology and Industry, Sun Dong, stated in a response to a legislative councilor's inquiry that the government has been proactively developing relevant digital infrastructure to seize the opportunities presented by the rapid advancement of artificial intelligence. The city's current aggregate computing power has reached 5000 petaFLOPS (PFLOPS), with the Cyberport AI Supercomputing Centre accounting for 60% of this capacity.
Furthermore, the Sha Ling Data Park, which has entered its construction phase, is projected to provide 180,000 PFLOPS of computing power by 2032. This represents a 36-fold increase over Hong Kong's current total capacity, offering robust support for AI computational needs, including token processing. Sun Dong explained that tokens are the fundamental units used by AI systems when processing text, commonly employed to gauge data processing volume and the associated computational power requirements.
The government will closely monitor the latest global and local developments in AI and continuously optimize various digital infrastructure projects. He also emphasized that while promoting innovative AI applications, it is crucial to effectively manage technological risks and ensure operations comply with safety and ethical standards.
The Office of the Government Chief Information Officer (OGCIO) has issued several key guidance documents, including the "AI Ethical Framework" and the "Hong Kong Generative AI Technology and Application Guide," which are applicable within the government and also serve as references for public organizations and the wider community. The government has established relevant norms for departmental AI application, personal data privacy protection, and information security.
In addition to these overarching guidelines, the OGCIO released the "Guide on Application of Artificial Intelligence for Government Departments" in February 2026. This guide mandates that content generated by AI tools, including videos and images, must have its authenticity verified by official or credible sources. It also prohibits inputting confidential or personal data into external or public AI platforms and forbids instructing AI to generate content that infringes on copyrights or intellectual property.
The "Government IT Security Policy and Guidelines" also set standards for account registration, management, and system operations. These include encrypting transmitted and stored data and requiring all departments to regularly conduct security risk assessments and audits for their information systems and related data.
To encourage departments to explore AI applications in a secure and controlled manner, optimize processes, and reduce the time and administrative costs associated with individual system procurement, the OGCIO has launched an "AI Solution Catalogue" and an AI Toolbox. These resources aggregate over 260 AI solutions that have undergone technical review and are suitable for government scenarios, such as text summarization, automatic translation, data analysis, and document drafting assistance. The catalogue specifies their applicable scenarios, technical architecture, and limitations.
Furthermore, the government is continuously enhancing talent development programs to improve civil servants' digital mindset and AI application capabilities. The OGCIO, in collaboration with the Civil Service College, has introduced tiered online AI training courses covering foundational knowledge, ethics and security, tool application, and even frontier technology trends and regional developments. Assessments are conducted to evaluate and reinforce learning outcomes. The office also regularly organizes leadership lecture series on "Innovation and Technology" to strengthen senior management's digital governance capabilities.
For the 2025/2026 fiscal year, the OGCIO arranged approximately 130 training courses related to digital technology, promoting the integration of theory and practice to cultivate a civil service proficient in leveraging technology. Both in guidelines and training, the government emphasizes the importance of "human-in-the-loop." Government departments must subject any AI-assisted content to manual review and verification by personnel before release to ensure its truthfulness and accuracy, and must never rely entirely on AI output.