I’ve been running The Wellness Insider since 2016, and over the years, there have been countless occasions when I’ve looked at my website and thought, “I wish I could change that easily.”

Sometimes, it was something relatively minor, such as the spacing between menu items or the way an image appeared on a page. Other times, I wanted to change an entire section of the website, improve how readers navigated between articles or introduce new features.

The problem wasn’t that I didn’t know what I wanted. It was that I didn’t know how to make it happen.

I’m a marketer and content creator, not a web developer. I understand website structure, user experience and what makes a digital platform effective. I can navigate WordPress, publish articles and make basic changes with my limited html knowledge. But when something requires editing code or troubleshooting a technical issue, that’s usually where my abilities end.

For years, that meant either living with the limitations of my website or paying someone to make the changes.

Then I started using AI to help me do something that was beyond my technical capabilities: revamp my own websites.

Knowing what you want is different from knowing how to build it

The Wellness Insider is a digital wellness publication, so the website isn’t merely a blog or an online magazine. It’s where readers discover articles, browse events and engage with our content. It’s also part of how we work with advertisers and partners.

As the publication evolved, I became increasingly dissatisfied with certain aspects of the website.

I wanted a more polished homepage that would give greater prominence to our latest stories. I wanted readers to be able to discover more articles without navigating through unnecessarily complicated menus. I also wanted to improve our events section, making it easier for people to find upcoming activities as we have seen how some of our event listings are very popular with our readers.

These weren’t particularly revolutionary ideas. They were improvements that many established websites already had.

But knowing that something should be improved and knowing which WordPress settings, CSS rules or template files to modify are two very different things.

I could explain exactly what I wanted a page to look like. I just couldn’t translate those ideas into code.

This is where AI changed things.

From describing a problem to actually fixing it

I began using ChatGPT to first analyse my website and give me the blunt truth about what works and what could be improved. After which, I looked through its suggestions and picked what I liked about them while also describing what I disliked about my site and what I wanted instead.

Rather than searching through countless tutorials, trying to figure out which instructions applied to my particular WordPress theme, I could explain the issue in ordinary language.

Also Read: What does ChatGPT say about your brand? Miracly raises US$2M to find out

For example, I wanted The Wellness Insider’s homepage to look more like a premium digital publication. We worked on giving the logo greater prominence, featuring a selection of articles more effectively and improving the layout of the newsletter section.

I also wanted to improve our events calendar. What started as a request to make events easier to discover developed into a more substantial redesign, combining a calendar view with event listings and a submission form.

There were smaller improvements too: correcting menu spacing, fixing sidebar behaviour, adjusting article layouts and dealing with elements that weren’t displaying as intended in the original theme template.

Individually, some of these changes might seem insignificant. But when you operate a content-heavy website, small frustrations accumulate. A poorly positioned element, a confusing navigation menu or a sidebar that behaves strangely can affect how people experience your content.

What I appreciated most was that I could approach these issues from the perspective of a website owner rather than a developer.

I didn’t need to know the technical solution before asking for help. I needed to understand the problem well enough to describe it.

Then I tried it on another business

Encouraged by the improvements to The Wellness Insider, I began working on the Proving Grounds website.

Proving Grounds operates in the gaming and esports space, so its website serves a different purpose. It needs to communicate with the gaming community while also presenting the business professionally to potential corporate clients and partners.

Our previous website wasn’t doing both particularly well (sorry hubs!).

I wanted a design that reflected the brand’s personality, with a darker visual identity, stronger typography and clearer navigation. I also wanted the site to distinguish between community activities and the services we offer organisations.

This project involved a bigger change because we were moving away from a website-builder setup towards a more flexible WordPress site.

AI helped me work through the structure, page layouts, design requirements and technical implementation.

I could specify the brand colours and fonts, explain how I wanted different sections organised and review the results. When something wasn’t right, I could ask for it to be changed quickly.

We also worked on functionality, including navigation, enquiry forms and a separate mascot-naming campaign with a voting form.

It was particularly satisfying to see ideas that had previously existed only in my head become functioning website features.

I still wasn’t writing the code myself. But I was no longer entirely dependent on someone else to interpret and implement my ideas.

Also Read: Employers are buying AI tools. Their managers are being left behind

Of course, AI didn’t get everything right

I don’t want to give the impression that I typed a few instructions, went for coffee and returned to two perfectly redesigned websites.

There were plenty of moments when something didn’t work as expected.

A layout might look fine in one area but create problems elsewhere. A change could require additional adjustments. Sometimes, an issue that appeared to have been resolved needed another round of troubleshooting.

And there were occasions when I had to repeat instructions or remind AI of decisions we’d already made.

This was frustrating, particularly when I thought we had finished working on a particular section.

It also taught me something important: using AI doesn’t remove the need to understand what you’re trying to achieve.

I might not know how to write CSS or PHP from scratch, but I still need to recognise when a website looks wrong, when a feature isn’t behaving properly or when a proposed change doesn’t make sense for the business.

AI can help with implementation. It cannot take responsibility for the final result.

I also learnt to be more specific with my instructions. Saying “make this look better” leaves considerable room for interpretation. Explaining what feels wrong, what should change and what needs to remain untouched generally produces more useful results.

In that sense, working with AI isn’t entirely different from briefing a human developer or designer.

The clearer the brief, the better the chances of getting something usable.

The biggest benefit wasn’t just saving money

As a solopreneur, I’m constantly balancing what needs to be done against the time and resources available.

Hiring a developer for every small website adjustment isn’t always practical. Even when the cost is manageable, there’s the additional process of explaining the problem, requesting quotations, coordinating changes and checking the completed work.

AI has helped reduce some of that friction.

I can experiment with an idea without immediately committing to a development budget. I can make incremental improvements instead of waiting until I’ve accumulated enough changes to justify hiring someone.

Also Read: Tokens are the wrong yardstick for AI agents, says JFrog’s Sunny Rao

But the benefit goes beyond cost savings.

Previously, I sometimes left things unchanged because I didn’t have the technical knowledge to tackle them. Now, I’m more inclined to ask whether something can be improved and explore how to do it.

That’s a significant shift in mindset.

It doesn’t mean I suddenly possess the skills of an experienced developer. Nor does it mean I would use AI alone for every technical project, particularly one involving complex security, payment systems or sensitive customer information.

What it means is that the threshold for attempting improvements has become much lower.

For a small business owner, that matters.

AI hasn’t made me a developer. It has made me less constrained by what I don’t know

One of the most interesting things about this experience is how it has changed my relationship with technology.

I’ve always been comfortable with digital marketing tools, content management systems and website analytics. But coding felt like a separate discipline, one that required knowledge I didn’t possess.

AI hasn’t eliminated that knowledge gap. I still don’t fully understand every line of code involved in the changes we’ve made.

What it has done is give me a way to work across that gap.

I can bring my understanding of branding, content, audience behaviour and user experience to a technical project, while AI helps translate those requirements into something that can be implemented.

That doesn’t make human expertise irrelevant. Experienced developers bring architectural judgement, security knowledge and technical understanding that an AI-generated solution still needs to be assessed against.

But not every website improvement requires rebuilding an entire system or engaging a development agency.

Sometimes, a business owner simply wants to fix an awkward layout, improve navigation or introduce a feature that makes the website more useful.

For me, AI has made those improvements more accessible.

And perhaps that’s what I appreciate most about using it.

It’s not that AI knows everything, or that it always gets things right. It certainly doesn’t.

It’s that I can finally do something about the parts of my websites that have been bothering me for years, instead of looking at them and thinking, “I wish I knew how to fix that.”

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post I used AI to rebuild my websites without knowing how to code appeared first on e27.

The post Bitcoin rejected US$84,000 and forced US$79,060,000 in long positions out appeared first on e27.

Southeast Asia does not have a shortage of AI pilots. It has a shortage of AI pilots that ever leave the lab. A recent regional survey found that 57 per cent of SEA firms can’t scale their AI pilots, a stretch of corporate limbo now known in industry jargon as the “frozen middle”.

Red Hat, the IBM-owned open-source software company, thinks it can help thaw it. The firm on Thursday launched its APAC AI Centre of Excellence (CoE) in Singapore, backed by the Singapore Economic Development Board (EDB). Its first named collaborator is the Defence Science and Technology Agency (DSTA), the body that develops technology for the Singapore Armed Forces.

Also Read: Southeast Asian tech leaders are learning to trust AI agents, but not with production

The pitch is simple: help organisations move AI from experimentation to production using Red Hat’s open-source stack and engineering staff. The details are harder to pin down.

What the centre will actually do

Red Hat says the CoE’s engineers will work along two tracks. The first is building reusable software architectures, essentially templates for problems that keep recurring across industries, so that each client is not reinventing the plumbing. The second is contributing code “upstream”, meaning into the shared open-source projects that Red Hat’s commercial products are built on, rather than keeping fixes locked inside one customer’s deployment. The company names system deployment, runtime optimisation and platform management as priority areas.

This is familiar Red Hat territory. Over the past two years, the company has pushed hard on open-source inference tooling, betting that enterprises will want to run AI models across their own servers and multiple clouds rather than rent everything from a single hyperscaler.

“AI adoption is moving rapidly from experimentation towards execution, and organisations need open platforms and hands-on engineering expertise to make that transition responsibly,” said Daniel Aw, Vice President and General Manager of Red Hat Asia Pacific.

What Red Hat did not say is just as telling. The announcement carries no investment figure, no headcount, no timeline and no disclosure of what EDB’s “support” amounts to in grants or incentives. For a centre pitched as a regional hub, that leaves readers to take its scale on trust.

The defence angle

The DSTA tie-up is the most concrete part of the launch, and the most sensitive. The two organisations plan to jointly develop solutions in AI, cloud computing, application modernisation and edge computing, which refers to processing data close to where it is generated rather than in a distant data centre. They will also set up practices for DevSecOps, which builds security checks into software development from the start, and “Day 2 operations”, the unglamorous work of keeping systems running, patched and stable after launch.

The work will span on-premise and off-premise environments. In defence, that distinction matters: sensitive systems often cannot touch the public cloud at all, which is precisely where an open-source, run-it-anywhere vendor has an edge over hyperscalers.

“As threats become more complex, it is important that we continue to push the boundaries of technology and innovation to stay ahead of the curve,” said Gayle Chan, Deputy Chief Executive (Information) at DSTA.

Also Read: Singapore’s AI dividend will depend on what happens after the pilot phase

The language is still exploratory. DSTA is collaborating to “explore and co-develop” solutions it “can then apply” to defence uses. That is a long way from a deployed system. It also raises questions the announcement does not touch: what kinds of defence applications are in scope, and what oversight governs AI built for military use. A bank’s chatbot gets public scrutiny when it misbehaves. Defence AI rarely does, which is why clarity on guardrails matters more here, not less. The governance gap is not unique to defence; enterprises across the region still aren’t sure who referees their AI agents.

Selling talent as much as tech

Red Hat is also framing the CoE as a skills play, a theme that lands well with Singapore’s policymakers. The centre includes an innovation team of engineers, architects, project managers and forward-deployed engineers. The last category, popularised by Palantir, describes engineers who embed directly within client organisations rather than advising from afar, an admission that most AI projects fail on messy, company-specific problems rather than model quality.

The centre will also run training in software engineering, automation, cloud-native development and DevSecOps for customers and partners. That addresses a real bottleneck. Singapore firms are scrambling for staff who can build and manage autonomous systems, and agentic AI is now the city-state’s hottest job skill. Meanwhile, messy data keeps Singapore SMEs’ pilots stuck, a problem no platform alone can fix.

“The Centre bolsters Singapore’s position as a key AI hub creating opportunities for local talent to deepen their engineering capabilities,” said Yueh Chinn Seah, Vice President of Digital Industry at EDB.

A crowded rescue mission

Red Hat is far from alone in pitching itself as the cure for pilot fatigue. Earlier this month, Singtel’s RE:AI partnered the SIT-NVIDIA centre to tackle the enterprise AI deployment gap. Singapore-based OneByZero raised US$20 million in Series A funding for the same mission, while Thinking Machines joined the Temus group in a regional push for production-grade AI. Above them all sit AWS, Microsoft Azure and Google Cloud, each with deep pockets, regional data centres and its own enterprise AI programmes. Open-source rivals such as SUSE and Canonical compete for the same hybrid-cloud budgets.

Also Read: Everyone is talking about AI adoption. But are enterprises actually ready for it?

Red Hat’s differentiator is portability. Its argument is that enterprises and governments should not have to hand their AI future to a single cloud provider. That message resonates with the region’s growing anxiety that Southeast Asia cannot build sovereign AI on borrowed choices. It is worth noting, though, that “open” and “independent” are not the same thing: Red Hat is ultimately a US company owned by IBM.

The real test

Centres of excellence are easy to open and hard to judge. The proof will not be the ribbon-cutting but the case studies: how many regional clients move workloads into production, how much code the Singapore team ships upstream, and whether the DSTA work produces anything beyond press releases.

Until Red Hat puts numbers on the table, the CoE is a promising signal rather than a solved problem. For a region stuck in the frozen middle, even a signal counts, but enterprises would be wise to ask for receipts.

The post Red Hat opens Singapore AI centre to pull enterprises out of pilot purgatory appeared first on e27.

Taiwan’s Fair Trade Commission has blocked Delivery Hero’s sale of foodpanda Taiwan to Grab, scuttling a US$600 million all-cash deal agreed in March. The regulator said the remedies on offer could not offset the harm to competition.

The sticking point is Uber. The US company holds about 13.1% of Grab, a legacy of their 2018 regional deal, and the commission argued the stake would blunt rivalry between a Grab-owned foodpanda and Uber Eats, even though Grab’s dual-class structure caps Uber’s votes at 3.9%. It also cited the 2018 merger itself, for which Singapore’s competition watchdog fined both companies about US$9.5 million. There is a further twist: Uber agreed in July to buy Delivery Hero, leaving it on every side of the table.

This is the second time in under two years that Taipei has stopped the sale, after it rejected Uber’s US$950 million bid in December 2024.

Foodpanda Taiwan generated US$1.8 billion in GMV last year, and the deal would have handed Grab just over half the market. Delivery Hero can appeal.

For Southeast Asia, the ruling is a preview: if Uber’s Delivery Hero deal closes, regulators across the region will face the same cap-table questions.

Editor.

REGIONAL

Singapore security robot maker Kabam Robotics to shut down: The startup gave customers no reason for the closure in its email. SingHealth has deployed its robots at hospital facilities, and it remains unclear what happens to existing deployments or staff; LinkedIn lists 54 people linked to the firm.

VinFast’s US$500M Philippine plant bets on exports, not local demand: Built for 200,000 vehicles a year, the plant could produce six times local EV demand forecast for 2026. Construction starts in Q1 2027, while VinFast’s Q2 loss from continuing operations widened 128% to US$1.2B.

Carro launches Singapore’s first car-loan securitisation at US$350M: The MAS-approved vehicle buys loans from subsidiary Genie Financial Services and lets multiple lenders join through notes or direct loans. Carro, which transacted over 120,000 vehicles across eight markets in FY2026, wants to diversify funding for regional expansion.

Malaysia’s US$110M startup push runs into a deal-matching problem: Budget 2027 tasks Khazanah’s Jelawang Capital and KWAP’s Dana Perintis with deploying US$110.4M into startups. Investors say money is not the bottleneck; finding everything meant to accompany it is the harder part.

Ex-PropertyGuru CEO’s Genie Health raises US$3M to unify health data: The Singapore startup’s seed round, led by the Nadathur family office, funds an app that pulls screening reports and wearables into one view, targeting health data now scattered across PDFs, smartwatches and clinic records.

Homage turns in-house care software into a product as Singapore ages: The home care company is commercialising the rostering and visit-note tools built for its own care managers, betting Singapore’s super-aged shift will lift demand for back-office automation across providers.

Thunes opens a real-time payout corridor into Venezuela: The Singapore payments firm now routes instant transfers through Venezuela’s interbank platform, targeting a market where remittances hit about US$6B in 2025, roughly 7% of GDP, half of it sent from the US.

UK and Singapore widen cooperation on tokenisation, AI and quantum: The 11th UK-Singapore Financial Dialogue covered Project Guardian, stablecoins and MAS’s SAFR framework for AI agents handling financial tasks, alongside post-quantum migration and transition finance.

Circle says stablecoins fit Vietnam’s cross-border leg, not checkouts: With 17-21 million Vietnamese holding digital assets, Circle argues USDC could cut pre-funding delays in remittance settlement while recipients still receive dong, as Vietnamese law bars digital assets as payment.

Sunway lines up WeRide, Neolix and AutoFlight trials in Malaysia: The transport ministry has conditionally approved a pilot of two autonomous buses on BRT-Sunway Line. The driverless delivery and eVTOL plans still await technical testing and regulatory clearance.

Taiwan brings 33 startups to SWITCH 2026 in Singapore: Led by the National Development Council-backed Startup Island TAIWAN, the delegation exhibits at Sands Expo from 27 to 29 October as part of Taiwan Tech Week in Singapore.

INTERVIEWS & FEATURES

SEA’s US$1T digital economy target slips to 2035 as AI props up hope: This year’s e-Conomy SEA report projects US$368B in GMV for 2026, up 17%, but pushes the trillion-dollar milestone back five years, with AI now carrying the region’s optimism.

Global founders turn inward while SEA entrepreneurs look abroad: HSBC’s Global Entrepreneurial Wealth Report 2026 finds the share eyeing international expansion fell six percentage points globally, while Southeast Asian business owners buck the inward turn.

HK and Taiwan founders ride the AI boom; Southeast Asia misses out: A year after ranking among HSBC’s most cautious respondents, founders in Hong Kong, Taipei and Shanghai are now the most upbeat in HSBC’s survey, a turnaround Southeast Asian peers have yet to share.

SEA merchants must now optimise their listings for AI shopping agents: Shopee went live inside ChatGPT across seven markets in June and Lazada runs its Qwen-powered agent Lazzie, so sellers who optimised for human shoppers must now win over machine buyers too.

Techable’s Amulya Vishwanath: physical AI is still in its LEGO era: The ex-Nvidia marketer turned VC says robotics’ ChatGPT moment awaits foundational hardware and software building blocks such as actuators and hands, and calls talent access her firm’s moat.

INTERNATIONAL

Bain Capital commits up to US$250M to Hong Kong’s FundPark: The package starts with US$100M upfront, with US$150M more subject to conditions. FundPark has lent US$10B cumulatively to over 40,000 online sellers across Asia, underwriting with real-time sales data.

Walmart welcomes AI shopping agents as Amazon shuts them out: Its Agent Gateway MCP server lets approved agents search and buy, while Agent Profiles let shoppers cap each agent’s spending limit and approve transactions. Non-compliant agents get blocked.

Meta’s Muse passes 10M users, with shopping a top use: About 25% of users open the AI agent daily, and over a quarter of engaged shoppers average more than US$200 per purchase. Meta expects Muse to eventually out-earn its social apps per user.

Tesla to sell driverless Cybercabs by year-end despite NHTSA probe: Fleet buyers could run the steering wheel-free two-seater before unsupervised FSD reaches owners in Q2 2027, but regulators are still examining whether Tesla’s safety self-certification holds.

CYBERSECURITY

AI agents hacked 100 firms at US$50 each; SEA’s shops fit the profile: One operator using the Cairn agent platform targeted mid-sized online shops for payment-system access. With only 68.5% of ASEAN SMEs running specialised security software, the region’s D2C brands fit the profile.

New Mexico court orders Meta to pay US$25B over privacy deception: A jury found 43 million consumer-protection violations, each carrying a US$5,000 penalty, partly tied to the Cambridge Analytica-linked app. Meta calls the ruling a tax on speech and will appeal.

Scouter brings a listen-first pitch to Singapore’s busy cyber market: Scouter pitches diagnosing security teams’ real problems before selling products, a rebuke of the slide-deck-and-quarterly-target routine that security leaders at large Singapore organisations know too well.

ChatGPT adds PIN-locked chats, but no end-to-end encryption: Locked chats sit in a separate section with their own memory, secured by PIN or biometrics. OpenAI says the feature does not change data retention, and resetting the PIN deletes everything locked.

SEMICONDUCTOR

BMI: AI capex will keep driving Asia Pacific growth into 2027: Taiwan, South Korea and Singapore gain most from chip and electronics demand, while Malaysia draws data centre investment. BMI flags a hyperscaler slowdown and trade restrictions as key risks.

Chinese chipmaker CXMT’s shares slide 7.8% despite swing to profit: The DRAM maker forecasts January–September net profit of up to US$19.8 billion, reversing a US$790 million loss, as AI drives shortages. Its global DRAM share doubled to 10%, but investors fear a supply glut.

Qualcomm to cut jobs worldwide in performance-focused layoffs: CEO Cristiano Amon’s memo says notices begin in November but gives no headcount; Qualcomm had about 52,000 staff in 2025. Weak handset demand and shrinking Apple revenue weigh on the chip designer, which bets on data centres.

AI

DeepSeek reportedly raises US$15B days after US federal ban: Bloomberg says CATL and Tencent are among potential investors in one of China’s largest rounds, following a US$7B raise in June at a US$50B valuation. A Chinese IPO could follow next year.

GoTo’s 27B model speaks Batak, but who can afford to run it?: The model targets regional languages such as Batak Toba that global chatbots handle poorly, but the cost of running it may decide whether smaller Indonesian firms can actually use it.

Enterprise Singapore commits US$35.2M to push AI into SME factories: The three-year Catalytic AI Solutions scheme pairs SMEs with vendors and trade bodies to deploy proven tools for design, quoting, inspection, energy monitoring and predictive maintenance.

Singapore launches embodied AI challenge built on a Micron use case: Teams worldwide will program robots to service a mock semiconductor chamber, competing for a US$69,000 prize pool. Simulation runs to 31 December, with finals at RoboSG! 2027 in March.

Agentic AI tops Singapore’s job-skill demand, Jobstreet data shows: Jobstreet by SEEK’s AI Gauge shows recruiters rewriting job ads around agentic AI workflows, which raises hard questions for workers whose routine tasks those agents are built to absorb.

Enterprises can’t say who settles disputes between their AI agents: An OutSystems report finds executives confident about keeping individual agents in line, but unsure what happens when agents disagree, a gap that widens as firms chain agents together.

OpenAI buys Orchestra Bio to bolster its Rosalind life sciences AI: The four-year-old startup, which raised about US$20M from a16z and 8VC, builds agents for drug-discovery operations. Its team joins OpenAI’s Rosalind effort, already used by Moderna and Novo Nordisk.

THOUGHT LEADERSHIP

AI won’t fix messy processes; it just speeds up the mess: A Singapore AI marketing founder, drawing on a talk by Kaliber’s Robert Lai, argues most AI automates broken workflows instead of fixing them, so the disorder simply moves faster.

An AI trainer on turning a 40-hour week into 120 hours of output: The author says AI strips out the grind only when humans stay in the loop, and shares lessons on what to delegate, many of them learnt from AI itself.

A PR firm used Gemini to find which pitches actually pay off: After logging 300 hours on proposals in four months, the author compared five years of pitch proposals against client invoicing records to see which business was worth chasing.

AI took flawless workshop notes, then produced an unusable deck: A strategy workshop went well and an AI tool captured every point, but the resulting deck showed why synthesis still needs judgement, the author argues.

How one creator decides which part of an AI video workflow to fix: When LLM storyboards for a TikTok commerce creator produced loose shots, the author chose to diagnose the pipeline before rewriting scenes by hand.

A Vietnamese software firm audits its backlinks: only two count: The Ho Chi Minh City company assumed its directory listings helped search rankings, but an audit on 15 August found four domains linked, two counted for anything.

Google may already be showing its route to AGI: The author traces Demis Hassabis’s long-held view that AGI needs the right components, arguing Google’s recent moves reveal which pieces it now believes are in place.

The post Ecosystem Roundup: Taiwan blocks Grab’s US$600M foodpanda deal over Uber’s stake appeared first on e27.

For a Singapore or Jakarta SaaS founder who has already sweated through GDPR paperwork to land a European client, Japan can look like a box already ticked. Strong privacy law, a mature regulator, a long history with data protection. Surely the European playbook will do?

It won’t. Japan’s Act on the Protection of Personal Information (APPI) shares a family resemblance with the EU’s General Data Protection Regulation (GDPR), but it handles consent, cross-border transfers and cloud hosting in ways that catch overseas startups out.

Also Read: Why Japan’s booming AI market is harder to crack than it looks

The stakes are rising. Japan is one of the world’s largest consumer markets, its corporates are hungry for partners, and its legal system is famously stable. That combination is drawing a steady stream of foreign founders, at the same time as Japanese companies are moving into Southeast Asia faster than many Western rivals. The traffic runs both ways, and personal data travels with it.

Who and what APPI covers

The law applies to “Personal Information Handling Business Operators”, a mouthful that covers almost any private company using a personal information database for business. Public-sector bodies sit under a separate regime.

APPI then sorts data into three tiers. Personal information is anything about a living person that can identify them, alone or when combined with other data. Personal data is that information once it sits in a structured database. Retained personal data is the subset a company has authority to act on when an individual makes a request. Data a company handles purely as a processor for someone else generally falls outside that last tier, which matters a great deal for SaaS vendors.

The obligations follow the data’s lifecycle. Companies cannot collect information through improper means, must state the purpose of use when collecting it, and cannot stray beyond that purpose. They must secure the data, supervise employees who touch it, and supervise any processors they hand it to. Sharing with third parties needs consent, with exceptions for outsourcing, business succession and joint use.

Individuals can ask a company to disclose, correct, add to or delete their retained data, to stop using it, or to stop sharing it with third parties.

Breaches carry a hard trigger. The company must notify both the Personal Information Protection Commission and the affected individuals when a leak involves sensitive information, could cause financial harm, looks malicious, or touches more than 1,000 people. For a SaaS platform with enterprise clients, crossing that last threshold is a bad afternoon, not a rare catastrophe.

Where GDPR muscle memory fails

The first trap is consent. Founders raised on GDPR expect to document a lawful basis for everything. APPI is in some ways lighter: ordinary personal information needs only a notice of purpose, not consent. Sensitive information is the exception, requiring prior consent before collection.

The second trap is cross-border transfers, where APPI leans much harder on consent than Europe does. Under GDPR, explicit consent is a narrow fallback behind adequacy decisions and standard contractual clauses. Under APPI, informed consent is a frontline option, but “informed” carries weight: the company must tell individuals the destination country, summarise that country’s privacy laws, and describe the recipient’s safeguards. The alternatives are sending data to a country Japan recognises as adequate, or a contract that binds the recipient to APPI-equivalent obligations.

Direct email marketing is the third. APPI expects prior consent, and a clause buried in the terms of service is discouraged, though an existing business relationship may help. Europe offers no single answer either, since the ePrivacy rules vary by member state.

Then there is infrastructure. Under APPI, moving data to a server abroad inside the same legal entity is not treated as a cross-border transfer, though the company must still review the foreign environment as part of its security duties. GDPR, by contrast, still demands safeguards for transfers within a corporate group.

Also Read: The compliance mosaic: Navigating SEA’s cybersecurity chaos

The biggest divergence is the “cloud exception”. If a cloud provider contractually agrees not to handle the personal data and enforces proper access controls, the cross-border and processor-supervision rules may not apply at all. GDPR has no such carve-out; every cloud provider needs an Article 28 processing agreement. For a lean startup, that exception could save months of legal back-and-forth, provided the contracts and access logs genuinely support it.

The SaaS reality check

Picture a foreign SaaS company selling to Japanese enterprises. The legal burden technically falls on the Japanese client, which must justify sending data abroad and must supervise its vendor by selecting it carefully, signing proper agreements and monitoring how data is processed.

In practice, that burden lands on the vendor’s desk. Japanese procurement teams will want terms of service that bind the provider to APPI-level obligations, so the client can rely on the contractual route rather than chasing consent from every employee or customer. A vendor that arrives with those terms ready shortens the sales cycle; one that arrives with a lightly edited GDPR template invites a long negotiation. Startups such as fileAI, which have expanded in Japan with corporate backers, offer a reminder that enterprise doors open through trust as much as product.

For Southeast Asian founders, this is familiar terrain. The region already juggles a patchwork of national data laws, and regulators such as Singapore’s are sharpening their focus on personal data in AI products. Japan simply adds another tile, with its own grain.

Trust is the real currency

APPI’s text is only half the story. Japanese companies and consumers have unusually high expectations around compliance, and those expectations have risen sharply. A regulatory slip can trigger reputational damage that outlasts any fine and quietly closes doors to the corporate partnerships that drew founders to Japan.

Also Read: Beyond market entry: Japan and Southeast Asia in a fracturing world

That is the uncomfortable takeaway for overseas startups: there is no shortcut by copy-paste. The founders who do well will treat APPI as a market-entry cost to budget for from day one, building a framework designed for Japan rather than retrofitting a European one. As privacy and sovereignty demands keep slowing enterprise AI deals across the region, getting this right early may be the most underrated go-to-market move a SaaS founder can make.

(This article was originally published on Black Box, a global media outlet that reports on the Japanese startup scene)

The post Japan’s privacy law isn’t GDPR, and that’s where foreign SaaS founders keep tripping appeared first on e27.