Generative AI Governance Principles (Hong Kong)
Tags: Frameworks
TL;DR
- Five principles: legal compliance; security and transparency; accuracy and reliability; fairness and objectivity; practicality and efficiency.
- Calls for explainability, content labelling, RAG-based grounding, fact-check tools, and diverse data.
Why it matters for HK marketers: These principles are what clients and regulators expect to see reflected in your AI workflows, disclosures, and vendor SLAs.
The principles
Compliance with laws and regulations
- Respect IP and personal privacy at all lifecycle stages; adhere to stricter overseas requirements when applicable; avoid disseminating false/harmful information.
Security and transparency
- Mitigate model/service risks; disclose risks to users; employ encryption and explainable AI; contrast transparency trade-offs of open-source vs proprietary models.
Accuracy and reliability
- Reduce hallucinations via RAG; provide user-friendly fact-check tools (authoritative retrieval, intelligent comparison); label AI content.
Fairness and objectivity
- Diversify data sources; prevent discriminatory outputs across belief, gender, age, ethnicity, income, etc.; avoid information silos.
Practicality and efficiency
- Optimise architectures and workflows to solve real problems across tasks and industries; improve speed, relevance, and quality.
5 key governance principles set in the Guideline.
So what for marketers
Bake these principles into creative reviews and platform selection; require vendors to evidence risk disclosures, grounding methods, and bias controls.
← Back to Knowledge Base