Module 4: AI Regulation · BoK IV.B
China, Japan and the rest of the world
China runs a multi-layered, use-case-specific network overseen by the CAC, requiring security reviews and algorithm registration. Japan takes innovation-first soft law via the AI Promotion Act - regulation by guidance with no penalties. Singapore's Model AI Governance Framework and AI Verify, plus Brazil, Canada, India and the UK, round out the global sweep.
Two contrasting Asian approaches, then the global sweep.
- 🇨🇳 China → multi-layered, use-case specific → no single unified law; use-case rules (Interim Measures for GenAI Services (2023), Algorithmic Recommendation Provisions, Deep Synthesis Provisions); the CAC oversees, exempting research institutions and requiring security reviews + algorithm registration where a service can influence public opinion or 'mobilize' the public.
- 🇯🇵 Japan → innovation-first soft law → the AI Promotion Act relies on business cooperation and current laws; government can investigate and advise, but no penalties; this is Japan's tradition of regulation by guidance over punitive enforcement. APAC pattern → Japan, Singapore and South Korea take a more relaxed approach than the EU for now.
| Where | Approach |
|---|---|
| 🇧🇷 Brazil | Proposed comprehensive risk-based AI bill → human oversight for high-risk AI; right to an explanation of decisions; three risk levels similar to the EU; clear damage rules |
| 🇨🇦 Canada | 2023 Voluntary Code of Conduct for advanced generative AI → fairness, safety, human oversight |
| 🇮🇳 India | Principles plus national committees; proposed Digital India Act replacing the IT Act of 2000; complements the DPDP Act; MeitY advisories require platforms not to facilitate unlawful content and untested models only after outputs are labelled fallible or unreliable |
| 🇸🇬 Singapore | Innovation-friendly → 2019 Model AI Governance Framework, Asia's first, voluntary; AI Verify testing toolkit; 2024 GenAI framework with nine dimensions |
| 🇬🇧 UK | No central AI law as of 2025 → sector-specific, flexible 'context-based' oversight from the 2023 AI Regulation White Paper |
| 🌍 Others | EU member states (Italy, Spain), Switzerland, Norway, Turkey; UAE/ADGM/DIFC Regulation 10; Kenya, Nigeria, South Africa strategies; Australia, Taiwan sectoral codes; UN draft resolution and OECD/UNESCO/G7 principles guide but are not legally binding |
Key terms - quick answers
What is “CAC”?
Cyberspace Administration of China; oversees AI services, requiring filings, security reviews and algorithm registration.
What is “Interim Measures for GenAI Services (2023)”?
China's rules for public generative AI; exempt research institutions, require security reviews and registration.
What is “Deep Synthesis Provisions”?
China's rules requiring labelling and watermarking of deep-synthesis (deepfake) AI outputs.
What is “AI Promotion Act”?
Japan's innovation-first soft-law framework; government can investigate and advise but imposes no penalties.