"I told the chatbot I'm 66 million years old. It still asked me to verify my age."
— Regional Chief-Dino, Asia
Introducing Toto-Rex
A dinosaur lost his job to an algorithm, so he set out to understand it: the rules, the regulations, and just enough of the technology to follow them. This edition follows Asia. Singapore, China, Hong Kong, India and Japan; in plain English, from credible sources. Serious about the learning. Less serious about himself.
The newsletter is coming soon.
No technical background. No shortcuts. Just one word, one rule, and one small victory at a time, written down so you can walk the same path without the late nights and cold coffee.
This is that path, tidied up and shared. Welcome in.
AI news from verified outlets across the globe. Asia first: SG, CN, HK, IN, JP.
The world's first dedicated rulebook for AI companions has been in force since 15 July 2026. China's Interim Measures on anthropomorphic interactive services require covered services to disclose that users are interacting with AI, watch for signs of dependency, intervene when a user signals self-harm, and apply strict safeguards for minors. Finalised in April by the CAC and four other ministries, the Measures apply to services offered to the public within mainland China that simulate human personality and communication through continuing emotional interaction. Ordinary non-emotional customer-service, knowledge-answering, work-assistant, education and research tools are expressly outside these Measures. Cross-border providers should obtain advice on the territorial facts. Separately, the AI-agent Implementation Opinions published on 8 May 2026 are a guiding policy instrument with no stated commencement date. Paragraph 11 directs regulators to use filing, testing and defective-product recall measures for regulator-defined sensitive and key-industry scenarios under applicable law; it is not a general 15 July duty for all agents.
Read it on cac.gov.cn →Phase two targets AI-fabricated news, face- and voice-swap impersonation, violent or vulgar output, harm to minors and AI-run troll accounts: 49,000 accounts penalised, 2,400 sites and apps dealt with.
Feedback is in on the draft guidelines that decide when scraping "publicly available" data can lawfully train a model. Final text expected within months.
Multi-agent guidance, automation-bias controls; and a companion paper asking who is legally responsible when an AI agent acts.
Eight years after FEAT, the central bank is consulting on guidelines for AI model risk management across banks, insurers and capital markets.
The Implementation Opinions, published on 8 May 2026, are a guiding policy instrument with no stated commencement date. Paragraph 11 directs regulators to determine open scenarios and, under applicable law, regulatory requirements and safety standards, use filing, testing and defective-product recall measures in sensitive sectors and key industries. It is not a general 15 July duty for all agents.
The region's biggest AI gathering brings labs, robots and ministers to the riverfront; with governance sharing the main stage again.
App stores and platforms keep pulling AI products that skip the mandatory visible-plus-hidden labels on synthetic content, in force since last September.
Universities, institutes and companies doing sensitive AI research are standing up the ethics committees the new review measures require.
The privacy regulator's compliance-check rounds continue, measuring real AI use against its Model Personal Data Protection Framework.
The Monetary Authority's sandbox lets banks trial generative AI under supervisory eyes before anything touches a customer.
The AI Research & Development Institute funded in the 2025 Budget moves from announcement to build-out, anchoring the city's AI ambitions.
Building on last year's FREE-AI framework, the central bank's draft guidelines would formalise model governance for banks and regulated lenders.
The IT Rules amendments make "synthetically generated information" a defined legal category; labelled, traceable, and taken down fast.
With the Rules notified last November, India's consent-first privacy regime phases in; and AI training data sits squarely inside it.
The machinery created by the AI Promotion Act; chaired by the Prime Minister; steers the national plan for promotion-first AI governance.
Nikkei, the Asahi and the Yomiuri press their copyright suits against the AI answer engine; the case testing Japan's famously training-friendly law.
Sell AI into Europe from Singapore, Shenzhen or Bengaluru and the transparency obligations have followed the product since 2 August; high-risk duties follow from 2 December 2027.
Headlines link to their original outlets; the one-line gists are Toto-Rex's own. Curated 22 September 2026; check the source for the latest.
Open a folder; the desk shows only what you need. Less scrolling, more finding.
The side rail and menus still reach everything; picking a link opens its folder for you.
Each item summarised by Toto-Rex, linked to its original source. Asia first: Singapore, China, Hong Kong, India, Japan.
How this works: every summary is Toto-Rex's own plain-English take, and each card links out to its original source so you can read it yourself. Nothing is reproduced here. Educational only, not legal advice — for your own situation, ask a suitably qualified professional.
Scope. This briefing covers AI compliance, regulation and policy for Singapore, mainland China, Hong Kong, India and Japan, plus the regional work (ASEAN, the Hiroshima Process) that binds them together. The events radar (section 05, below) covers those five jurisdictions and online events from recognised hosts. It does not try to cover the US or Europe; the UK & EU edition does that.
Sources. Credible, verifiable sources only: the regulators themselves (IMDA, the PDPC, the CAC, the PCPD, MeitY, METI and the PPC), established international and Asian law firms, recognised legal press, and peer-reviewed academic journals. Every card links to its primary source.
Depth & timeline. Summaries are short by design; what a thing is and why it matters, then off you go to the original. Current as of 27 September 2026. China's binding human-like AI Measures have applied since 15 July 2026, and the CAC's "Qinglang" AI clean-up campaign reported its second-phase results on 2 September; its separate AI-agent Opinions remain policy guidance without a commencement date; India's DPDP Rules phase in through 2027; Singapore's advisory guidelines on generative AI and personal data sit alongside the PDPA rather than replacing it.
Limitations. Toto-Rex is a dinosaur, not your lawyer. Summaries are his interpretation, they simplify, and much of this law is written in Chinese or Japanese; where an English summary and the official text differ, the official text prevails. Check the live source before you act, and take advice for your own circumstances.
Six numbered desks; one per jurisdiction, plus the cross-border play; each expandable when you want the detail.
What this is: the working map of how each jurisdiction actually governs AI, in the order a product usually meets them; based on the official sources linked in each desk, as at 7 July 2026. Rules and dates change; the links are the truth. Educational only, not legal advice — for your own launch, brief counsel in each market.
One region, five philosophies; not five drafts of the same law. Read straight down: each line is one question, answered across all five jurisdictions.
Toto-Rex's note: build one governance programme to the strictest common denominator. China's filings, the labels China and India now demand, and Singapore's insistence on a named accountable human; then trim it down per market. Never the other way around.
Beyond these five. South Korea's AI Basic Act (Asia's first comprehensive statute, in force January 2026), Taiwan and the wider ASEAN bloc are moving too. This edition holds to the five above by design.
Toto-Rex's note: the agentic framework launched at Davos in January 2026 with a companion paper asking the question of the decade; who is legally responsible when an AI agent acts? Singapore publishes its homework; everyone else copies it.
Toto-Rex's note: a unified AI Law keeps being proposed and keeps being postponed; the spring 2026 line from state media was that conditions are "not yet ripe". The application-by-application chain is the system. Plan for it, not around it.
Toto-Rex's note: Hong Kong regulates AI the way it regulates most things; existing law, sector regulators, and the strong suggestion that you would rather not be the example. The HK$1bn AI R&D Institute says the ambition is real.
Toto-Rex's note: the India–AI Impact Summit in New Delhi (February 2026) put a Global-South stamp on the agenda; adoption and impact first, with the New Delhi Declaration to show for it. India writes rules with teeth where it hurts: impersonation, elections, money.
Toto-Rex's note: Tokyo's courts have settled one thing already. DABUS lost here too. An inventor must be human (Tokyo District Court 2024, upheld 2025). The machine does the work; a person signs the form.
Toto-Rex's note: he keeps the five questions on one card above the desk. Asked before launch they are a checklist; asked after launch they are an incident report.
Sources. Every rule and date above links to its official source. IMDA, the PDPC, the CAC (with established English translations), the PCPD, MeitY and IndiaAI, Japan's Cabinet Office and METI; last updated 7 July 2026.
Scope. General business use of AI across Singapore, mainland China, China (Hong Kong SAR), India and Japan. Not covered: tax, employment, sector licensing, export controls, and anything bespoke to your situation.
Limitations. Toto-Rex is a dinosaur, not your lawyer; and much of this law is written in Chinese or Japanese, so summaries simplify and translations can lag. Always check the linked source, and brief counsel in each market before you act.
One more time, because it matters: everything on this desk is education, not legal advice — laws change and your facts are yours alone. Consult your lawyers before you act. Toto-Rex will still be here when you get back.
Fresh AI uses in the wild; and the rule that walks in behind each one.
This week's front page belongs to the machines that feel like people; and the agents that act like staff. On 15 July China's binding Measures for anthropomorphic AI interactive services take effect: covered companion services must disclose the machine behind the voice and meet safeguards for dependency, minors and crisis signals. China's separate AI-agent Implementation Opinions were published on 8 May with no stated commencement date; they direct regulator-led filing, testing and defective-product recall measures for defined sensitive and key-industry scenarios under applicable law, not a general 15 July duty for all agents. Singapore answered the agent question months earlier with a voluntary framework. Toto-Rex's working rule travels across all five jurisdictions: give an agent the least access it needs, log every step it takes, and keep a human hand on anything that moves money or people. Autonomy transfers work. It never transfers responsibility.
— T.R., from the study corner
How this note works: each morning Toto-Rex picks one focus from the front page (finance, health, agents, hiring) and writes its compliance angle in plain English. The focus rotates; the discipline doesn't. Educational only, never legal advice.
Assistants now complete multi-step tasks (browsing, form-filling, booking) through "computer use" controls. Marvellous, and also a liability surface, because an agent's click is your click. The rule that follows: Singapore's Model AI Governance Framework for Agentic AI (the world's first) says bound the risk, keep humans accountable, control the tools. China's agent Opinions direct regulators to use filing, testing and defective-product recall measures for defined sensitive and key-industry scenarios under applicable law, but do not create a general 15 July duty for all agents.
IMDA · the agentic frameworkFrom radiology triage to national screening programmes, clinical AI is deployed and scaling across Asia's hospitals. The rule that follows: health ministries treat it as regulated medical technology. Singapore's MOH publishes dedicated AI-in-healthcare guidelines, and each jurisdiction's device regulator stands behind them.
MOH SingaporeWuhan runs one of the world's largest driverless-taxi fleets, and more cities license services every quarter. The rule that follows: city-by-city pilot permits, safety-driver phase-outs, and liability rules that keep the operator; not the passenger; answerable.
Baidu Apollo · official siteFrom 15 July, covered services must disclose that users are interacting with AI, build anti-addiction safeguards, intervene on self-harm signals, and apply strict protections for minors. The Measures are limited to public services within mainland China that provide continuing emotional interaction; ordinary non-emotional assistants are expressly excluded. The rule that follows: classify the service before assuming this specialist rulebook applies, then check the other CAC regimes separately.
CAC · the MeasuresApplications change weekly; the desk keeps only what is verifiably deployed, linked to its official source. Educational only; not legal advice, and never a product endorsement.
Markets, funding and the rules that referee them; refreshed daily.
Meta has signed its first AI datacentre deal in India with Reliance, planting hyperscale AI infrastructure on the subcontinent; while Johor, Batam and Japan's regional grids fill with the same concrete. The models make headlines; the buildings make balance sheets. For the region's regulators the questions follow the money: power, water, data residency, and who answers when the workloads are sovereign.
Read it on TechCrunchSingapore's central bank wrote Asia's first AI-in-finance principles (FEAT, 2018), built the Veritas assessment methodology with the industry, and is now consulting on formal guidelines for AI model risk management across banks, insurers and capital-markets firms. If you deploy AI in Singaporean finance, this is your rulebook's front door.
MAS · official siteThe FREE-AI committee's 2025 report set the framework for responsible AI in Indian finance; the RBI's 2026 draft guidelines would harden it into supervisory expectations; model inventories, validation, human oversight and board accountability for banks and regulated lenders.
RBI · official siteHong Kong's banking regulator pairs firm expectations; senior-management accountability, consumer-protection circulars, model risk management; with a generative-AI sandbox at Cyberport where banks trial use-cases under supervisory eyes. The SFC applies the same logic to licensed firms.
HKMA · official siteAsian exporters meet the EU AI Act the moment their systems or outputs reach European users: transparency duties since 2 August 2026, high-risk obligations from 2 December 2027 after the Omnibus deferral. The lesson for a Singapore fintech or an Indian SaaS firm is the same; the market decides which law applies.
EUR-Lex · the ActToto-Rex's standing reminder: this desk reports the money; it does not manage yours. Nothing here is investment, legal or tax advice — markets move faster than dinosaurs, so verify at the source and consult your own advisers.
Where AI meets law, policy and product; straight from the boards that matter.
The state that writes Asia's most-copied AI frameworks hires the people who write them; policy, standards, testing and the agentic-AI desk.
One desk, five regimes: PDPA, PIPL, PDPO, DPDP and the APPI. The IAPP's CIPP/A and AIGP are the badges recruiters search for.
MAS, the HKMA and the RBI are all raising the bar on AI model risk; and every institution in scope is hiring the people to meet it.
The frontier labs now hire in Asia; alignment, policy and trust & safety alongside engineering, in Tokyo and Singapore offices.
India's IT majors and thousand-plus capability centres are standing up AI governance teams as the DPDP Act and deepfake rules land.
Compliance checks, framework drafting, sandbox supervision; the region's regulators are staffing their AI desks. Public-service pay, front-row seat.
Every filing regime creates a profession: security assessments, algorithm registrations, labelling audits and content compliance for China's AI products.
Toto-Rex re-pins this board every morning from the official sources linked; postings, salaries and closing dates change daily, so always verify on the board itself. No placement fees, no affiliations; he just likes seeing dinosaurs employed.
AI summits, conferences and seminars from credible hosts. Singapore, China, Hong Kong, India, Japan and online.
If you leave the study corner for one thing this year, make it this. SWITCH is Singapore's flagship innovation festival and the front door to the whole Asian tech scene; tens of thousands of founders, investors, researchers and regulators, government-backed and thoroughly international. AI runs through every hall, IMDA and the national AI programme use its stages for the announcements that become next year's frameworks, and it lands in the same city and season as the Singapore FinTech Festival; one trip, two summits.
Why this one, and why you can trust it: Toto-Rex took no fee, no ticket and no commission to name it; this is simply the most useful, most popular and most prestigious innovation gathering on the Asian calendar. Read on for the full radar below.
Official site & registrationDates, venues and prices change; check them out yourselves before you go. Anything free is marked.
Real cases, real fines, one fictional lake; and your rights at every turn.
Pick a jurisdiction. Each opens with one featured overview — Toto-Rex's plain-English summary of how that place governs AI; then at least ten notes on the real cases, fines and rules beneath it: what actually happened, what the rules say, what a company should never do, and what you can do about it. Every piece states its scope and links its sources, and none of it is legal advice.
How to read these: the cases, fines and dates are real and each note links its sources so you can verify them yourself; please do. The lake, Nessie and Colin are entirely fictional, and clearly labelled as such. None of this is legal advice, and it does not create any professional relationship; for your own situation, consult your lawyers. Based on the sources as at 7 July 2026; the law moves quickly.
The library. These notes are the public face of Toto-Rex's AI Compliance Library; at version 6.1.1, 187 research documents (about 115,000 words): twelve deep jurisdiction packs (UK, EU, US, Singapore, Hong Kong, mainland China, India, Japan, South Korea, Brazil, Canada, Russia), an EU-27 implementation matrix, a US 50-state & D.C. index, a wider regional atlas, and the policy and evidence templates beneath them. Research cut-off: 13 July 2026.
How it is sourced. The English research edition records 431 unique official-source URLs across 1,160 citations and 2,610 internal links. The library is not re-cut every time this page changes: anything added after the cut-off carries its own last-updated date. Automated structural and citation checks are useful quality controls, but they are not legal or linguistic certification.
What it is not. The library's own label applies here too: general information and customisable templates — not legal advice or certification. English is the controlling edition; official legal texts and their authoritative language versions prevail.
Just enough of the technology to understand the compliance. No coding required.
Fifteen guided tracks from absolute beginner to specialist; including the five-jurisdiction Asia primer. Step-by-step lessons, graded quizzes, four-depth study notes, flashcards, and a printable certificate when you pass. Every claim carries its source.
Certificates will be a paid option, launching soon.
Toto-Rex is not paid to feature any of these, and there are no affiliate links on this site. Listings point to each provider's official page; details, prices and dates change, so check before you enrol. Anything free is marked.
One term, in plain English. No maths. No jargon.
Toto-Rex's note: always check what an AI tells you against a proper source. Especially names, numbers, and anything you would put your own name to. It is the cheapest insurance you will ever buy.
8–12 terms from the full glossary, refreshed daily. Ignore spaces and hyphens; letters only. Toto-Rex is watching. Supportively.
Talks, interviews and podcasts; builders, regulators, researchers and critics, in their own words.
Singapore's CNA covers the region's AI build-out better than anyone with a camera: the model labs of Hangzhou, the datacentres of Johor and Mumbai, the regulators writing the rulebooks; documentary-grade reporting, free on its verified channel, from the newsroom closest to the story.
Watch on CNA's channelThese are external shows Toto-Rex rates. Links go to each show's official page; specific episodes come and go, so browse for the guest you want.
Toto-Rex wants to try them all. Forgive him; he's a dinosaur, picking it up bone by bone.
Grouped by job. Listings link to each tool's official site; none of it is paid placement, and most have a free tier.
Rights, duties, penalties, dated agendas, operational practice and cross-border reach.
Five expert dossiers most relevant to this edition are available here in compact English form. The interactive globe, all fourteen jurisdictions, treaties, comparative models and industry tracker remain in the Global edition.
Everything this site teaches, bound, sourced and indexed.
AI has arrived faster than the law that governs it; and that gap is where careers, companies and rights are now won or lost. The book strips away the maths, the hype and the fear: the EU AI Act, the GDPR and UK data protection in plain English, built around a practical six-factor framework you can run on any AI system; five factors that follow its life, and a sixth you hold against them all. Written for founders, lawyers, educators, regulators and the professionally curious.
"A working machine asks: who answers for me?". Like everything in the study corner, the book is education, not legal advice. It sets out the law, cases, news and studies as at September 2026, reminds you to check them against the primary sources, and then tells you what the current situation is.
A monthly-ish letter; he aims for once a month, study schedule permitting, and would rather skip an issue than send you filler. Toto-Rex writes it himself, between study sessions and waiting on deliveries. The developments that matter, what the new rules mean, and one tool worth a look. Plain English, no jargon, no commitment.
Sign-up opens soon.
Cheers. See you next month. Probably.
The real thing is still in the workshop.
Toto-Rex is coming to phones. The waitlist opens soon.
Short films, songs about AI and honest tool tests; serious learning, carried lightly.
No technical degree here, and no rush; just one exam at a time. All the new AI can feel overwhelming, and a touch of FOMO is only natural, but once you know what you're aiming for, it's far easier to find your way through. Your pace, your path.
“Who says a T-rex can't reach their toes?” — Toto-Rex, stretching
Everything above is the serious stuff. This is the cherry on top: useless facts, small joys, and a dinosaur with opinions.
He told everyone the "T" in T-Rex stands for Technology. We both know it stands for Tyrannosaurus.
A 25 cm dome of solid bone on its head, evolved for headbutting rivals. And, Toto-Rex suspects, recruitment software.
Roughly how long his lineage survived. He still cannot get a CV past an automated screening filter.
A healthy attitude to a backlog, or a coping mechanism. Possibly both.
Of the Scottish Highlands. Also named Nessie, she runs a flourishing marketing company on the shores of Loch Ness; and is entirely out of his league, which, he notes, has never stopped anyone on a dating app.
Distant relations, technically. They hang about a certain museum looking down on everyone, and they haven't kept up with a single development since the Cretaceous. He visits anyway. Family is family.
The one thing on this site that is not negotiable.
A machine can draft, sort, total and summarise. It cannot care, keep a promise, own a mistake, or read a room. That part of the work was never in the job description. It was always you.
So learn the tools and mind the rules. Then bring what no model can produce: taste, judgement, kindness, and the nerve to ask the obvious question.
“Whatever the algorithm decides, you've got me, buddy.”
Toto-Rex
“The difference between AI and a person is that a person can feel pain. Then again, what would I know? I’m just a dino boy who eats fish.”
Step away from the syllabus. The dinosaur plays a patient game.
He reads more than he lets on.
He may or may not read this. He is usually studying, or waiting on a delivery. But he reads more than he lets on, and he appreciates you stopping by.
The postbox opens soon.
Toto-Rex can make mistakes. Everything here is general information, not legal advice: check the primary source, and take advice from a qualified professional before you act.
