Twelve Ethical AI Principles (Hong Kong)
Tags: Frameworks
TL;DR
- Twelve principles guide responsible AI, with two “Performance Principles” required to enable the rest.
- Principles derive from human rights and Hong Kong ordinances, including PDPO.
Why it matters for HK marketers: These principles define the minimum bar for trustworthy AI in customer-facing and data-driven marketing.
Structure of the principles
- Performance Principles (must-have foundation)
- Transparency and Explainability: Ability to explain AI decision-making clearly to humans.
- Reliability, Robustness and Security: Long-term reliable operation using appropriate models/data; robustness to errors; security against cyber-attacks.
- Fairness: No unjust bias or harm; treat similar individuals similarly.
- Diversity and Inclusion: Respect and consider interests of all impacted stakeholders.
- Human Oversight: Degree of human intervention scales with severity of ethical issues.
- Lawfulness and Compliance: Adhere to laws, regulations, and regulatory regimes.
- Data Privacy: Align with PDPO DPP1–DPP5 on collection, accuracy/retention, purpose/use, security, and openness.
- Safety: Do not compromise physical safety or mental integrity.
- Accountability: Clearly identifiable accountable party for use and misuse.
- Beneficial AI: Promote the common good.
- Cooperation and Openness: Foster multi-stakeholder cooperation.
- Sustainability and Just Transition: Mitigate societal and environmental impacts.
Implementation notes
- Performance Principles enable validation of other principles (e.g., without reliability/security, compliance to other principles can’t be verified).
- Data Privacy references PDPO’s Data Protection Principles explicitly for collection, use, and safeguards.
12 principles total; 2 are foundational Performance Principles.
So what for marketers
Operationalise these principles in briefs, model documentation, and vendor contracts; prioritise explainability and reliability first to make the rest auditable.
← Back to Knowledge Base