Hong Kong: Guidance on Ethical Development and Use of AI

Latest developments

On 18 August 2021, the PCPD issued the Guidance on Ethical Development and Use of AI (AI Guidance) to help organisations understand and comply with the relevant requirements of the Personal Data (Privacy) Ordinance (PDPO) when developing or using AI, and to self-assess whether practices recommended in the AI Guidance have been adopted. Notably, the ambit of the PCPD in relation to AI is only relevant to the extent that an AI system processes personal data.

Summary

Data Stewardship Values and Ethical Principles for AI

The AI Guidance recommends three fundamental Data Stewardship Values, namely, being respectful of the dignity, autonomy rights, interests and reasonable expectations of individuals; beneficial and fair to stakeholders when organisations develop and use AI. In line with international standards such as the European Commission’s Proposal for Regulation of AI and UNESCO and OECD Recommendations on AI, the AI Guidance sets out the following seven ethical principles for AI:

• Accountability: Organisations should be accountable for what they do and be able to provide sound justifications for their actions;
• Human Oversight: Organisations should ensure that appropriate human oversight is in place for the operation of AI where AI system users should be informed and able to act autonomously with regards to recommendations or decisions of the AI systems;
• Transparency and Interpretability: Organisations should disclose their use of AI and relevant policies while striving to improve the interpretability (i.e. the ability to determine the cause and effect) of automated decisions and decisions made with the assistance of AI;
• Data Privacy: Effective data governance should be put in place to ensure proper handling and protection of personal data involved in the development and use of AI;
• Fairness: Organisations should avoid bias and discrimination in the use of AI – any differential treatment between different individuals (or groups of individuals) should be validly justifiable;
• Beneficial AI: Organisations should use AI in a manner that benefits and minimises harm to stakeholders; and
• Reliability, Robustness and Security: Organisations should ensure reliable operation of AI systems which are resilient to errors and guarded against attacks.

Practice Guide

The practice guide within the AI Guidance provides practical examples of how organisations should approach AI governance when implementing AI in their operations, from inception to implementation and ongoing risk-based management, covering the following areas:

• AI strategy and governance;
• Risk assessment and human oversight;
• Development of AI models and management of overall AI Systems; and
• Communication and engagement with stakeholders.

How could it be relevant for you?

Organisations that intend to or have begun to develop and use AI technology in their operations are advised to assess the risk levels of their AI systems and the data privacy concerns involved and to implement the PCPD’s recommended best practices for better compliance with the relevant requirements of the Personal Data (Privacy) Ordinance.

Next steps

With Hong Kong’s vision of becoming a regional data hub in innovation and technology and the increasing applications and use of AI technology in business, it is anticipated that there will be more robust AI development and governance in Hong Kong.

*Information is accurate up to 27 November 2023