AI Policy Tracker
AI regulations and policy briefs by country
🌍 Country Policy Guide In-depth AI policy for 15+ countriesG7 Releases Common AI Governance Framework, Advances Mutual Standards Recognition
G7 summit reaches five AI governance consensus points: 1) Cross-border AI risk information sharing mechanism; 2) Mutual recognition of AI safety testing standards; 3) Coordinated gen-AI content labeling norms; 4) Prevention of AI misuse in military domains; 5) Ensuring AI benefits are accessible to developing nations. Japan as 2026 G7 President made AI governance a core agenda item.
OECD Updates AI Principles with Generative AI Transparency Requirements
OECD releases 2026 AI Principles with new generative AI chapter: mandatory training data disclosure for high-risk AI; cross-border safety incident reporting standards; minimum labeling for AI-generated content. 38 member countries commit to domestic legislation by 2027.
China "AI+" Strategy Upgrade: ¥100B Computing Subsidies and Open-Source Model Support
Following global attention on DeepSeek and other domestic models, China rolls out systematic AI support: 1) NDRC establishes ¥100B AI computing subsidy fund targeting SME costs; 2) Ministry of Science launches "AI+" integration plan for manufacturing, healthcare, education; 3) MIIT announces 50 inaugural AI demonstration zones; 4) Encourages open-source foundation models and establishes international AI cooperation whitelist.
South Korea's AI Basic Act Enters into Force
South Korea's AI Act enters force — Asia's first systematic AI basic law. Requires pre-deployment safety assessments for high-impact AI, establishes AI incident liability mechanisms, and creates dedicated AI Committee for policy coordination.
OpenAI Releases GPT-5: Massive Reasoning Leap, Native Multimodal Surpasses GPT-4o
OpenAI officially releases GPT-5, outperforming predecessors across benchmarks: 1) Math reasoning (MATH-500) score up to 92% (vs. GPT-4o's 77%); 2) Code generation (HumanEval++) pass rate reaches 88%; 3) Native video input with real-time processing; 4) Context window expanded to 2M tokens; 5) Two subscription tiers: Plus ($20/mo) and Pro ($200/mo), with API access for enterprises. Analysts predict GPT-5 will accelerate AI adoption in healthcare, legal, and education sectors.
Anthropic Releases Claude 4 Series: Opus 4 Reasoning Matches GPT-5, Safety Alignment Leads
Anthropic simultaneously releases three Claude 4 models — Opus 4, Sonnet 4.6, Haiku 4.5: 1) Opus 4 matches GPT-5 in scientific reasoning, long-document analysis, and complex coding; 2) 'Extended Thinking' mode makes reasoning process transparent and auditable; 3) Sonnet 4.6 is the developer-preferred performance/cost balance; 4) Constitutional AI (CAI) 2.0 framework debuts — harmful instruction refusal accuracy up to 99.2%; 5) Claude Code has become the most-used AI coding agent among professional developers, integrated with VS Code and JetBrains.
China Issues National AI Foundation Model Standards, Accelerates AI Infrastructure
China's SAMR releases first national AI foundation model standards governing training data, safety evaluations, and output labeling. Concurrently, NDRC approves 50 new computing hubs targeting 300 EFLOPS domestic intelligent computing capacity by 2027.
EU AI Act Prohibited Uses Apply; GPAI Model Rules Take Effect
EU AI Act core clauses phase in: prohibited uses (social scoring, remote biometric surveillance) now apply; GPAI model transparency and copyright obligations effective. Companies must complete high-risk AI compliance by August 2026 or face €35M / 3% turnover fines.
Trump Revokes Biden AI Order, Shifts to "AI Dominance" Strategy
Trump administration revokes Biden's 2023 AI executive order and issues new 'Maintaining U.S. AI Leadership' order — prioritizing reduced regulatory burden, accelerated government AI procurement, and U.S. dominance in global AI competition. Focus shifts from safety assessment to competitiveness.
US Stargate AI Infrastructure Initiative: $500 Billion to Reshape Global Compute Landscape
The White House announces the Stargate AI infrastructure investment initiative: 1) SoftBank, OpenAI, and Oracle commit to $500 billion over 4 years ($100 billion immediate) for US AI data centers and compute infrastructure; 2) First data center campuses in Texas expected to create 100,000+ jobs; 3) NVIDIA to supply hundreds of thousands of H100/H200 GPUs; Intel and AMD also involved; 4) Initiative aims to maintain long-term US AI compute advantage over China; 5) Implications for multinationals: AI compute costs declining long-term, but geopolitical risks are accelerating global compute supply chain restructuring.
UK Hosts Global AI Safety Summit
Bletchley Park AI Safety Summit achieves Bletchley Declaration with 28 countries committing to cooperate on AI risks and establishing first inter-governmental AI safety network.
EU AI Act Enters into Force
World's first comprehensive AI regulation enters force, imposing strict oversight on high-risk AI systems with fines up to 7% of global annual turnover.
Singapore Updates AI Governance Framework
Singapore launches AI Governance Framework 2.0 with new generative AI governance guidelines, providing actionable compliance checklists and toolkits for enterprises.
Japan Releases AI Business Guidelines
Japan releases AI governance guidelines for businesses, emphasizing soft-law governance and encouraging voluntary adoption of best practices rather than mandatory regulation.
US Executive Order on Safe AI Development
Biden administration signs executive order requiring AI developers to share safety test results with government and establishes AI safety standards.
China Interim Measures on Generative AI Services
China's first regulation on generative AI takes effect, requiring service providers to comply with core socialist values and conduct algorithm filing and safety assessments.