AI GUIDE FOR BOARDS IN SINGAPORE
Section III: Governance and Oversight
AI-Related Regulations, Frameworks and Tools: Examples
• China: Measures for Labelling AI-Generated Content Mandates that all network service providers structurally watermark and clearly label text, video, audio, or digital humans generated via generative AI (including strict labelling mandates for livestreaming e-commerce virtual hosts). • China: AI Safety Governance Framework 2.0 National parameters dictating safe data laundering, negative data sampling, and adversarial training rules to stop algorithmic “data poisoning” before model commercialisation. • Hong Kong: Model Personal Data Protection Framework for AI A high-level regulatory baseline tracking procurement, internal lifecycle gates, and routine risk assessments to prevent breaches of the statutory Personal Data (Privacy) Ordinance. • Hong Kong: High-Level Principles on AI Twelve structural directives explicitly telling banks how to structure governance when outsourcing models to third parties, preserving human-in-the-loop overrides for consumer lending decisions. • Australia: National AI Plan A tiered framework targeting speci c, high- impact deployment elds (e.g., automated recruitment, critical infrastructure, law enforcement) with mandatory safety standards and algorithmic transparency protocols. • ASEAN: ASEAN Guide on AI Governance and Ethics (including Gen AI Supplement) A multi-country blueprint specifying seven pillars for regional interoperability: Transparency, Fairness, Security, Robustness, Human-centricity, Privacy and Accountability.
• Singapore: Model AI Governance Framework for Agentic AI
Expanded to explicitly target autonomous AI agents capable of reasoning and independent planning. It mandates upfront risk bounding, strict checkpoints for human accountability, and technical logging to combat automation bias (the over-trusting of autonomous systems). • Singapore: Advisory Guidelines on Use of Personal Data in AI Enforceable guidelines under the Personal Data Protection Act governing how organisations collect and clean data for algorithmic training and recommendation engines. • Singapore: AI Verify Testing Framework e AI Verify Testing Framework helps organisations assess the responsible implementation of their AI systems against 11 internationally recognised AI governance principles. e Framework is also aligned with other international frameworks such as those from the EU, G7, OECD and the US. • China: Amended Cybersecurity Law is cornerstone framework explicitly embeds AI safety into national defence. It expands extraterritorial enforcement capabilities and allows for massive corporate nes (up to 5 per cent of turnover or up to RMB10 million) for signi cant data leaks or critical system failures caused by AI. • China: Measures for Administration of AI Anthropomorphic Interactive Services A strict regulation taking eect targeting AI companion apps, virtual partners, and conversational bots. It explicitly bans virtual intimate relationships for minors, mandates automatic “minor mode” cutos, forces a pop- up alert a er 2 hours of continuous use to prevent emotional dependency, and prohibits generating material related to self-harm.
AI governance frameworks and tools aligned with internationally-recognised principles can help boards navigate cross-border complexities.
56
Made with FlippingBook Digital Proposal Creator