Hong Kong AI Ethics Principles
Tags: Frameworks
TL;DR —
- A set of 12 principles underpinning ethical AI adoption in Hong Kong.
- Two are foundational “Performance Principles”: Transparency & Explainability, and Reliability, Robustness & Security.
- Data Privacy aligns with PDPO’s Data Protection Principles (DPP1–DPP5).
Why it matters for HK marketers: These are the non-negotiables brands must evidence when using AI for targeting, personalization, and automation in Hong Kong.
Structure of the principles
- Performance Principles (must-have foundation):
- Transparency and Explainability — Organisations must be able to explain AI decision-making in clear, comprehensible terms.
- Reliability, Robustness and Security — AI must operate reliably over time, be resilient to errors and adversarial attacks, and meet security requirements.
- Fairness — Treat similar individuals fairly, without discrimination or harm.
- Diversity and Inclusion — Promote inclusive and diverse usership; consider all stakeholder interests.
- Human Oversight — Calibrate human intervention to the severity of ethical issues.
- Lawfulness and Compliance — Act in accordance with applicable laws and regulatory regimes.
- Data Privacy — Reflects PDPO DPP1–DPP5: lawful/fair collection with limited purpose; accuracy/retention limits; use limitations; security; and openness about policies and data held.
- Safety — Do not compromise physical safety or mental integrity.
- Accountability — A clearly identifiable, accountable party for implications and misuse.
- Beneficial AI — Promote the common good.
- Cooperation and Openness — Foster multi-stakeholder open cooperation.
- Sustainability and Just Transition — Mitigate societal and environmental impacts.
How they are used in the framework
- Embedded throughout the AI Practice Guide and tested via the AI Application Impact Assessment.
- The Performance Principles enable verification that the other principles are consistently upheld.
12 Ethical AI Principles.
2 Performance Principles underpin 10 General Principles.
So what for marketers
Run every AI use case through these principles: ensure explainability, prove model reliability/security, and document privacy compliance to PDPO DPPs before launch.
← Back to Knowledge Base