AI & DATA REGULATION · EIGHT JURISDICTIONS
If your organization touches data or AI across borders, you're living inside four questions: What protects the data? What governs the AI? What controls the transfer? And where must the data physically stay? This brief answers all four across the EU, UK, US, China, Canada, Brazil, India, and South Korea — every answer sourced, tiered, and dated.
Edition 2 · verified through 28 Aug 2026 · 23 pagesA report about data should be data. Download the full PDF — and the complete verified dataset behind it, machine-readable for your compliance team, your spreadsheet, or your AI agents.
Every jurisdiction is read through the same four questions, and every answer carries its governing instrument, its status, four separate dates, and a source tier — because blending "enacted," "in force," "rules effective," and "when we looked" is the most common way a regulatory claim goes quietly wrong.
The settled layer — but "settled" doesn't mean still. Korea's new 10%-of-turnover penalty tier, the UK's DUAA amendments, and India's phased DPDP runway all land inside this edition.
The in-force comprehensive-statute cohort is smaller than the headlines suggest — the EU's phased AI Act and Korea's Framework Act lead it, while China binds hardest without an omnibus law at all.
Where regimes diverge hardest: the EU's adequacy web, China's authorization-and-threshold model, the US DOJ Data Security Program, and India's enacted-but-not-yet-operative blacklist.
Narrower than feared — mostly sectoral. And where we couldn't support a clean answer, the dataset says so explicitly: two cells ship as honest published gaps, not confident-sounding guesses.
Already-fixed dates between now and 2028 — none of them predictions; each read from the governing instrument, dated, and sourced in the dataset:
Every jurisdiction-dimension record ships with its status, governing instrument, four separate dates, source tier, confidence rating, and verification date — as JSON and CSV. The files carry their own machine-readable notice, so an AI agent that finds them learns the same thing a human reader does: this is general information, dated, and no substitute for counsel.
The least-certain readings are flagged on a public watchlist. Edition 2 publishes its own revision record rather than silently overwriting Edition 1. Primary instruments were read where it counts — official gazettes, EUR-Lex, and government legal portals. You don't have to take our word for any claim — the dataset tells you exactly where each one came from.