Hong Kong - AI Development Guidelines (2025)

Overview

The Hong Kong Generative Artificial Intelligence Technical and Application Guideline (the “Guideline”) was officially released by the Digital Policy Office (DPO) of the Hong Kong Special Administrative Region Government on April 15, 2025. This landmark document provides a comprehensive set of instructions and practical operational guidance for various stakeholders involved in the generative Artificial Intelligence (AI) ecosystem, including technology developers, service providers, and service users (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; Digital Policy Office, https://www.digitalpolicy.gov.hk/en/our_work/data_governance/policies_standards/ethical_ai_framework/; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The Guideline’s primary objective is to strike a critical balance between fostering innovation in AI technology and ensuring its responsible application, thereby constructing a robust governance framework specifically tailored to the unique characteristics and needs of Hong Kong (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). It addresses the multifaceted scope and inherent limitations of generative AI applications, elucidating potential risks such as data leakage, model bias, and errors that require proactive mitigation strategies (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The development of this Guideline was a collaborative effort, with the DPO commissioning the Hong Kong Generative AI Research and Development Center (HKGAI) to conduct extensive research, gather industry feedback, and propose appropriate codes and guidelines based on practical applications and international best practices (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). Commissioner for Digital Policy, Mr. Tony Wong, emphasized that the government's aspiration is for the Guideline to enable the industry and the public to develop and apply generative AI technology in a safe and responsible manner, while simultaneously encouraging innovative applications and fostering widespread adoption in Hong Kong (news.gov.hk).

Definitions

While the Guideline itself does not provide an exhaustive glossary of terms in the extracted information, it delineates the roles of key stakeholders and implicitly defines the scope of generative AI. The Guideline is tailored for three main types of stakeholders (Digital Policy Office):

• Technology Developers: These are entities or individuals responsible for creating, training, and maintaining the foundational models and algorithms that power generative AI systems (Digital Policy Office).
• Service Providers: This category encompasses organizations, including platform providers, that deploy generative AI technologies in customer-facing applications, acting as intermediaries between developers and end users (Digital Policy Office).
• Service Users: These are individuals or entities, including content creators and content disseminators, who utilize generative AI services for personal or professional purposes (Digital Policy Office).

The Guideline also implicitly addresses concepts such as 'generative AI technology,' which refers to AI systems capable of producing multimodal content like text, code, images, audio, and video, often leveraging key technologies such as latent space learning, probabilistic modeling, and generative strategies (Digital Policy Office). Furthermore, it refers to 'high-risk content' such as deepfakes or financial documents, which necessitate specific identifiers to ensure tamper-proof integrity (Digital Policy Office).

Governance and Institutional Framework

The primary governmental body responsible for the promulgation and ongoing oversight of the Hong Kong Generative Artificial Intelligence Technical and Application Guideline is the Digital Policy Office (DPO) of the Hong Kong Special Administrative Region Government (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The DPO is committed to continuously monitoring the latest technological and application developments in generative AI and has pledged to regularly update the content of the Guideline for reference by all sectors (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). Integral to the development of the Guideline was the Hong Kong Generative AI Research and Development Center (HKGAI), which was established in 2023 with funding from the AIR@InnoHK initiative (news.gov.hk). The HKGAI played a pivotal role by systematically summarizing practical insights from R&D processes, studying local AI applications, synthesizing global best practices, and providing recommendations on the accuracy, accountability, and information security of generative AI technologies and practices (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The DPO maintains close collaboration with HKGAI, relevant academic institutions, and industry groups to ensure the Guideline remains current and effective (news.gov.hk). The Guideline establishes a comprehensive Generative AI governance framework designed to promote the effective and beneficial use of GenAI (Digital Policy Office). Under this framework, all stakeholders are expected to clearly define the scope of their actions and diligently assess potential risks corresponding to the identified governance dimensions (Digital Policy Office). This includes recommendations for establishing dedicated data oversight teams within organizations to manage data responsibly and ensure compliance (Digital Policy Office).

Key Focus Areas

• Personal Data Privacy: The Guideline places significant emphasis on the protection of personal data privacy, urging stakeholders to ensure the safeguarding of user data. This includes adherence to the Personal Data (Privacy) Ordinance (Cap. 486) and establishing robust privacy protection measures (Digital Policy Office). Technology developers must ensure data collection respects privacy protections, and service providers are expected to establish user consent mechanisms (Digital Policy Office).
• Intellectual Property: Addressing concerns related to intellectual property (IP) in the context of AI-generated content, the Guideline mandates that data collection and the creation of AI-generated content must respect existing intellectual property laws and not contravene IP rights (Digital Policy Office).
• Crime Prevention: The Guideline aims to mitigate risks associated with the potential misuse of generative AI technologies, emphasizing the importance of preventing activities that could lead to crime or unethical behavior. Service users are specifically advised to avoid illegal and unethical conduct (Digital Policy Office).
• Reliability and Trustworthiness: A core principle is to ensure the accuracy, robustness, and overall trustworthiness of AI systems and their outputs. This involves addressing issues like model bias and errors, and promoting transparency and interpretability in AI decision-making processes (Digital Policy Office).
• System Security: The Guideline mandates the implementation of robust security measures, including secure development and design practices, to ensure the reliability and stability of generative AI technology. It focuses on preventing technical risks such as data leakage and ensuring the secure operation of services, especially when handling user data and generative content (Digital Policy Office).
• Fairness and Objectivity: In line with international best practices, the Guideline promotes principles of fairness and objectivity in the design, development, and deployment of generative AI systems, aiming to prevent discrimination and ensure equitable outcomes (Digital Policy Office).
• Human Oversight: The Guideline implicitly advocates for human oversight in AI applications to prevent ethical issues and allow for intervention where necessary (Digital Policy Office).

Implementation Framework

The Guideline provides practical, actionable advice tailored to the distinct roles and responsibilities of each stakeholder group within the generative AI ecosystem (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html).

For Technology Developers, the Guideline encourages the establishment of comprehensive and dedicated teams, such as data teams and quality control teams, to ensure compliance with applicable laws and regulations, manage data responsibly, and prevent discrimination (Digital Policy Office). They are urged to implement ethical model development practices, safeguard user data through robust protocols, and ensure thorough testing of AI systems (Digital Policy Office).

Service Providers are expected to develop a responsible generative AI service framework (Digital Policy Office). Their responsibilities include maintaining content integrity, ensuring robust privacy protection for users, and establishing clear user consent mechanisms (Digital Policy Office). Furthermore, they must pay close attention to the compliance and security of their services, particularly when handling user data and generative content, and are provided with suggestions to ensure the secure operation of their services (Digital Policy Office).

Service Users are encouraged to engage with generative AI ethically, diligently verify the outputs generated by AI systems, and thoroughly understand the potential risks associated with their use (Digital Policy Office). They are advised to avoid illegal and unethical behavior during use and to learn how to protect their personal privacy and data security (Digital Policy Office). The Guideline also offers advice on using technology services safely and measures for identifying and addressing potential security threats (Digital Policy Office). To support compliance across all stakeholder groups, the Guideline recommends embedding traceability mechanisms into AI-generated outputs, conducting regular audits, and implementing user feedback loops. For high-risk content, such as deepfakes or sensitive financial documents, the Guideline advises the inclusion of tamper-proof identifiers (Digital Policy Office). Independent evaluations are also recommended at all stages of development and deployment to ensure ongoing adherence to ethical and technical standards (Digital Policy Office).

Monitoring and Evaluation

The Digital Policy Office (DPO) is committed to an ongoing process of monitoring and evaluation to ensure the continued relevance and effectiveness of the Hong Kong Generative Artificial Intelligence Technical and Application Guideline (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). This involves a proactive approach to reviewing the latest technological advancements and application developments related to generative AI (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The DPO has explicitly stated its intention to regularly update the content of the Guideline, ensuring it remains current and responsive to the evolving landscape of AI technology and its applications (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). This continuous review process is facilitated by ongoing collaboration with the Hong Kong Generative AI Research and Development Center (HKGAI), as well as engagement with other relevant academic institutions and industry groups (news.gov.hk). The Guideline itself includes recommendations for stakeholders to implement internal monitoring mechanisms, such as conducting regular audits of AI systems and processes (Digital Policy Office). Furthermore, it encourages the establishment of user feedback loops to gather insights on the practical application and impact of generative AI, allowing for iterative improvements and risk identification (Digital Policy Office). Independent evaluations are also advised at all stages of development and deployment to ensure an objective assessment of compliance and performance (Digital Policy Office). These measures collectively aim to ensure that generative AI technologies are deployed and utilized in a manner that maximizes societal benefits while effectively minimizing associated risks.

Penalties, Liability, and Appeals

As the Hong Kong Generative Artificial Intelligence Technical and Application Guideline functions as a framework of best practices and operational guidance rather than a statutory law, it does not directly stipulate specific criminal penalties, civil penalties, or fines for non-compliance within its text. However, the Guideline explicitly states that the development, deployment, and application of generative AI technology in Hong Kong must strictly adhere to all relevant existing laws, policies, and regulatory standards (Digital Policy Office). Consequently, any activities or outcomes arising from the use of generative AI that contravene these underlying legal instruments would be subject to their respective enforcement mechanisms, penalties, and liability provisions. For instance, violations related to personal data privacy would fall under the Personal Data (Privacy) Ordinance (Cap. 486), which carries its own set of penalties for non-compliance (Digital Policy Office). Similarly, infringements of intellectual property rights, engagement in fraudulent activities, or dissemination of harmful content would be addressed by existing laws pertaining to copyright, crime, and public morality. The Guideline emphasizes that service users must avoid illegal and unethical behavior, implicitly warning against actions that could lead to legal repercussions (Digital Policy Office). While the Guideline itself does not establish a specific appeals process, any legal challenges or appeals would be handled through the established judicial system in Hong Kong, based on the specific laws that have been violated. The DPO's approach reinforces Hong Kong's commitment to a 'light-touch regulatory approach' to AI, relying on its robust existing legal framework to address challenges and disputes related to AI systems (Digital Policy Office).

Relationship to Other Instruments

The Hong Kong Generative Artificial Intelligence Technical and Application Guideline is designed to operate in conjunction with and complement Hong Kong’s broader regulatory and ethical landscape for technology and data governance. It explicitly requires strict adherence to all relevant laws, policies, and regulatory standards (Digital Policy Office). Key related instruments include:

• Personal Data (Privacy) Ordinance (Cap. 486): The Guideline underscores the importance of personal data privacy, mandating compliance with the PDPO, which provides a comprehensive legal framework for the collection, holding, processing, and use of personal data in Hong Kong (Digital Policy Office).
• Ethical Artificial Intelligence Framework: This Guideline builds upon and references the broader Ethical AI Framework previously developed by the Digital Policy Office, which provides guiding principles, leading practices, and an AI Assessment for government bureaux/departments and is also applicable to other organizations (Digital Policy Office).
• Government's Principles of Data Governance: The Guideline aligns with the overarching principles of data governance established by the Hong Kong Government, which prioritize secure data management and responsible use of information (Digital Policy Office).
• Policy Statement on Responsible Application of Artificial Intelligence in the Financial Market (October 2024): Issued by the Financial Services and Treasury Bureau, this policy statement focuses on promoting AI development in the financial sector while mitigating risks related to cybersecurity, intellectual property rights, and data privacy (Digital Policy Office).
• Consumer Protection in respect of Use of Generative Artificial Intelligence (August 2024): This circular from the Hong Kong Monetary Authority (HKMA) provides guiding principles for consumer protection in customer-facing applications utilizing GenAI (Digital Policy Office).
• Security & Futures Commission of Hong Kong Circular to Licensed Corporations - Use of Generative AI Language Models (November 2024): This circular addresses the use of generative AI language models by licensed corporations in relation to their regulated activities (Digital Policy Office).
• Guidelines on the Use of Generative Artificial Intelligence for Judges and Judicial Officers and Support Staff of the Hong Kong Judiciary: The Hong Kong Judiciary has also issued its own guidelines, indicating a cross-sectoral recognition of the need for responsible AI use (Digital Policy Office).

This layered approach signifies Hong Kong's strategy to integrate AI governance within its existing, robust legal and ethical framework, rather than creating entirely new overarching legislation for AI.

International Alignment

The Hong Kong Generative Artificial Intelligence Technical and Application Guideline demonstrates a clear commitment to aligning with international best practices and standards in AI governance. The development process itself, led by the Hong Kong Generative AI Research and Development Center (HKGAI), involved studying industry feedback and synthesizing global best practices (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html). The Guideline's key governance principles, which include security and transparency, accuracy and reliability, and fairness and objectivity, are explicitly stated to be in line with international practices (Digital Policy Office). Specifically, it draws reference from prominent international frameworks such as the OECD AI Principles and the principles underlying the European Union's AI Act (Digital Policy Office). This alignment ensures that Hong Kong's approach to generative AI governance is not developed in isolation but is harmonized with global efforts to address the ethical and technical challenges posed by advanced AI technologies. By drawing on these international benchmarks, Hong Kong aims to foster a trusted and responsible environment for AI innovation that is recognized and respected globally, facilitating cross-border collaboration and maintaining its reputation as an international financial and innovation hub. The continuous monitoring and regular updates by the Digital Policy Office are also intended to keep pace with evolving international standards and technological developments (Source: Digital Policy Office, https://www.digitalpolicy.gov.hk/filemanager/digitalpolicy/en/our_work/data_governance/policies_standards/ethical_ai_framework/HK_Generative_AI_Technical_and_Application_Guideline_EN.pdf; news.gov.hk, https://www.news.gov.hk/eng/2025/04/15/20250415mi014.html).

Implementation Timeline

Compliance Checklist

Sources and References