Hong Kong Generative AI Technical and Application Guideline and Ethical AI Framework

What it is

Hong Kong’s Digital Policy Office (DPO) publishes two companion documents on AI.

The Ethical Artificial Intelligence Framework was originally developed to help government bureaux and departments plan, design and implement AI and big data analytics in their IT projects. It consists of guiding principles, leading practices and an AI Assessment. The current edition, version 2.0 of December 2025, is revised for general reference by any organisation adopting AI. Its main components are twelve Ethical AI Principles, an AI governance structure, an AI lifecycle, an AI Practice Guide and the AI Application Impact Assessment template.

The Hong Kong Generative Artificial Intelligence Technical and Application Guideline was released on 15 April 2025. The DPO commissioned the Hong Kong Generative AI Research and Development Center (HKGAI) to research it, then promulgated the guideline having considered HKGAI’s findings. Version 1.1 was issued in December 2025, and the DPO has said it will update the guideline regularly.

Who it applies to

The Ethical AI Framework is written first for Hong Kong Government bureaux and departments, and states that its principles, practices and assessment template also apply to other organisations in general. The Generative AI Guideline is addressed to three groups, namely technology developers, service providers and service users of generative AI, across all sectors.

Neither document is legally binding. Binding obligations come from existing law, particularly the Personal Data (Privacy) Ordinance, which the Privacy Commissioner’s AI Model Personal Data Protection Framework explains for organisations using AI.

What it requires

Twelve Ethical AI Principles. Two are Performance Principles, Transparency and Explainability, and Reliability, Robustness and Security. The framework treats them as the foundation for the rest, because without them nobody can verify that the other principles are being followed. The ten General Principles are Fairness, Diversity and Inclusion, Human Oversight, Lawfulness and Compliance, Data Privacy, Safety, Accountability, Beneficial AI, Cooperation and Openness, and Sustainability and Just Transition.

AI Application Impact Assessment. The framework’s assessment template guides an organisation through the benefits and risks of each AI application and identifies the controls it needs. It has four components.

• Risk gating criteria, which set the level of assessment the application needs.
• Assessment questions across the AI lifecycle, each with a core theme and explanatory text.
• References to the AI Practice Guide for further guidance.
• Impact considerations, identifying beneficial and negative effects on specific stakeholders and the controls required.

The assessment works as a live document. It should be reviewed at four key stages of the AI lifecycle, with a copy kept at each stage, and repeated regularly (for example annually) or when major changes take place.

Generative AI governance. The Generative AI Guideline sets out the technical limitations and service risks of generative AI and the role of human oversight across the model lifecycle. It defines five dimensions of governance (personal data privacy, intellectual property, crime prevention, reliability and trustworthiness, and system security) and five key governance principles (compliance with laws and regulations, security and transparency, accuracy and reliability, fairness and objectivity, and practicality and efficiency). It then gives practical guidance to each audience. Developers are asked to build comprehensive teams and adopt proper working practices, service providers to build responsible service frameworks and development processes, and service users to act as proactive stewards of the technology.

Financial services. The Hong Kong Monetary Authority (HKMA) issued guiding principles on consumer protection in the use of big data analytics and AI on 5 November 2019. Its circular of 19 August 2024 to all authorized institutions extends that consumer protection approach to customer-facing uses of generative AI.

Key dates

• 5 November 2019. HKMA guiding principles on big data analytics and AI.
• 19 August 2024. HKMA circular on consumer protection in respect of the use of generative AI.
• 15 April 2025. Generative AI Technical and Application Guideline released.
• December 2025. Generative AI Guideline version 1.1 and Ethical AI Framework version 2.0.

How it relates to other frameworks

The Ethical AI Framework’s AI Application Impact Assessment is the nearest Hong Kong equivalent to an AI impact assessment under ISO/IEC 42005 or the Canadian Algorithmic Impact Assessment, and it pairs well with the risk assessment the PCPD framework asks for. The PCPD covers the personal data side, and the DPO covers the wider ethical and technical side.

The Generative AI Guideline surveys governance requirements elsewhere, including Mainland China’s generative AI rules, the EU AI Act, and Singapore’s Model AI Governance Framework, and draws on them for a Hong Kong approach. Organisations already working with Singapore’s generative AI framework or NIST AI 600-1 will find the five governance dimensions familiar.

A note on this page

This is Aicura’s reading of these documents, written to help you understand them. It is not legal advice and it is not the guidance itself. Read the primary documents from the Digital Policy Office and the HKMA, and take advice on how Hong Kong law applies to your organisation.