One region, five answers
Lesson 1 of 5 in Asia-Pacific: Japan, South Korea, Singapore, India, and Australia.
Between December 2024 and February 2026, the Asia-Pacific quietly became the most instructive laboratory in AI governance. Five advanced democracies faced the same technology, the same summits, the same OECD principles — and reached five different answers. South Korea passed a comprehensive statute. Japan passed a law with no penalties at all. Singapore built testing tools instead of rules. India regulated one narrow slice — synthetic content — through existing intermediary law. Australia studied mandatory guardrails for a year, then decided existing law was enough.
For a governance professional this region is a masterclass in a single question: how much law does AI need, and when? Every position on the hard-law–soft-law spectrum you learned in the soft-law module is occupied here by a real jurisdiction with real reasons.
South Korea
The statute. The AI Basic Act (Framework Act on AI Development and Establishment of Trust) passed 26 December 2024 and took effect 22 January 2026 — the world’s second comprehensive horizontal AI law after the EU AI Act. Risk-based duties for high-impact AI, generative-AI labeling, safety obligations for large models, extraterritorial reach, and fines. Lead regulator: MSIT. The deepest treatment in this module — the next lesson is all Korea.
Japan
The promotion law. The AI Promotion Act (2025) is a basic law that promotes AI research and use, creates a PM-chaired AI Strategy Headquarters, and imposes a duty to cooperate — with zero penalties or fines. It codifies Japan’s long-standing agile-governance philosophy: steer by guidance and reputation, not sanction. Backed by the METI/MIC AI Guidelines for Business (April 2024) and one of the world’s most permissive copyright regimes for AI training (Art 30-4).
Singapore
The toolkit. No AI statute, and no plan for one. Instead: the Model AI Governance Framework (2019/2020), a dedicated Generative AI framework (May 2024), and a family of practical instruments — AI Verify (testing framework + toolkit), Project Moonshot (LLM red-teaming), and the Global AI Assurance Pilot. Regulators IMDA and PDPC position Singapore as the place where AI governance becomes testable engineering practice.
India
The targeted rule. No standalone AI law — but a hard rule where it hurt: the IT Rules amendment of 10 February 2026 (effective 20 February 2026) obliges platforms to label synthetically generated content with permanently embedded metadata, after a bruising deepfake election cycle. Around it: the ₹10,000-crore-plus IndiaAI Mission, an AI Safety Institute (January 2025), the 2025 AI Governance Guidelines, and host of the AI Impact Summit (February 2026) — pro-innovation positioning with guardrails added surgically.
Australia
The existing-law bet. After publishing a Voluntary AI Safety Standard (ten guardrails, 2024) and a proposals paper for mandatory high-risk guardrails (September 2024), the government decided in late 2025 not to enact a dedicated AI act — relying instead on existing law (Privacy Act, Consumer Law, Online Safety Act) plus a National AI Capability Plan. An Australian AI Safety Institute was announced 25 November 2025. The counterweight: privacy reform is adding automated-decision transparency duties.
Key terms: horizontal regulation, hard law, soft law, high impact ai, agile governance
Tool: Global Governance Atlas — Open the Governance Atlas and click through Japan, Korea, Singapore, India, and Australia — compare instruments, dates, and status badges as you work through this module.
Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.