AIHK releases governance guidelines focused on agentic AI, recommends human approval before high-risk actions

AI-assisted source summary · not independently verified. Content standards

The Hong Kong Artificial Intelligence Association (AIHK) has released voluntary AI governance guidelines focused on agentic AI, recommending human approval before agents take hard-to-reverse actions such as initiating payments, making contractual commitments or sending external communications. The guidelines recommend classifying AI applications into three risk tiers, with organizations advised to classify AI agents connected to internal systems as at least higher risk. The four-stage roadmap starts with foundational governance measures that can be implemented within 90 days and applies to organizations of different sizes and across different sectors.