I’ve watched the same movie play out too many times: a management team receives a penetration testing report, sees a wall of findings with scary-sounding names, and immediately assumes their platform is on fire.

It’s not on fire. It’s almost never on fire.

But the report sure makes it look like it is. And that’s the problem I keep running into, not with the platforms, but with how people read these reports.

The translation problem

When you’re a CTO or an external CTO-as-a-Service advisor, part of the job is translating between the world of security tooling and the world of business decision-making. These two worlds speak very different languages. Security tools speak in volumes of automated findings. Business leaders speak in risk, cost, and “should I be worried right now?”

That gap is where the panic starts.

Over the years I’ve learned to get ahead of it. Before every VAPT (Vulnerability Assessment and Penetration Testing) cycle, I walk my clients through what to expect from the results, what the findings actually mean, what’s noise, and what deserves real attention. It’s part education, part expectation management, and part gentle reminder that a 200-page PDF full of findings does not mean the sky is falling. Sometimes it just means the scanner was very thorough and a bit too enthusiastic.

The goal is simple: give non-technical stakeholders the mental framework to read a VAPT report without losing sleep. Because the report is only half of the story. The other half, the part that actually matters, is interpreting those findings in the context of your platform, your architecture, and your specific business requirements.

The anatomy of a VAPT report (for humans)

Here’s what most people don’t realise about penetration testing: the raw output of any engagement is never the final verdict on your security. It’s a starting point for analysis.

VAPT teams rely on automated scanning tools to generate their initial findings. These tools are designed to cast an absurdly wide net. They flag anything that could theoretically be a concern. And I mean anything. Your OAuth integration with Google? Flagged. Your CDN serving static assets from a different domain? Flagged. A cookie that JavaScript can access because your entire framework was literally designed that way? You better believe that’s flagged. Any open port on the server, even port 80 or 443? Yup, also flagged.

This isn’t a flaw in the process. It’s how the process works. The tools are doing their job. The question is what happens next.

Also Read: Should cybersecurity be nationalised?

The quality gap nobody talks about

And here’s where it gets interesting.

Not all VAPT teams are created equal. In fact, there’s a pretty dramatic quality spectrum, and where your team falls on it determines whether you receive a useful, contextualised security assessment or a PDF-shaped anxiety attack.

Budget-oriented teams tend to optimise for volume. They run the tools, collect the output, and forward everything to the client with minimal filtering. The result? A report with dozens, sometimes hundreds, of findings, many of which are informational noise or outright false positives. It looks impressive. It fills a lot of pages. But it creates exactly the kind of alarm that derails productive conversations about actual security.

I’ve seen reports where the same exact finding was listed separately for every URL on the platform. Same issue, same root cause, same “vulnerability”, just presented 147 times to make the PDF thicker.

More experienced teams, and yes, they typically cost more, invest significant effort in triaging their tool output before presenting it. They separate signal from noise. They tell you what actually matters and why. They cross-reference previous engagement results instead of re-investigating known behaviours from scratch. Their reports are shorter, more accurate, and infinitely more useful. You’re paying for judgment, not just scanning hours.

Severity levels: A quick decoder ring

Every VAPT report categorises findings by severity. Here’s the practical translation:

• Critical and High. Stop what you’re doing and fix these. These represent real, exploitable vulnerabilities. In a well-maintained platform with regular dependency updates, strong authentication, and proper encryption, these should be rare. If your report is full of them, you have a genuine problem. If it has zero, congratulations. That’s the goal.
• Medium and Low. Read these with a calm mind. They often represent theoretical risks, hardening suggestions, or configuration preferences. Many are informational. Think of them as a security consultant saying “you could also do this” rather than “your house is currently on fire.”
• Informational. These are diagnostic notes. They describe how your platform behaves. They don’t indicate risk. You can acknowledge them and move on.

The number of findings in a report tells you almost nothing about how secure your platform is. A report with 150 findings and zero criticals is a dramatically better result than one with five findings and two criticals.

Also Read: Singapore’s cybersecurity paradox: Leading in digital, lagging in defense

False positives: The uninvited guests

Every, and I mean every, VAPT engagement produces false positives. These are findings that automated tools flag as potential issues but which, upon analysis, turn out to be expected framework behaviours, design decisions, or artefacts of the cloud infrastructure itself.

In a recent engagement, we documented over 20 false positives across two reports. The cloud provider’s own security infrastructure was triggering alerts during the scan. The scanning tools were essentially detecting the host’s defence systems and reporting them as application vulnerabilities. That’s like a home inspector flagging your alarm system as a security risk. Technically, something happened. Practically, it’s the opposite of a problem.

Context is everything

If there’s one thing I want people to take away from this, it’s this: a VAPT report must always be read in the context of the specific platform it was conducted against.

Security is not a one-size-fits-all discipline. A finding that represents a genuine vulnerability on one platform could be an intentional design decision on another. Session tokens in URLs? Alarming, unless they’re part of a standard OAuth handshake with a provider like Google or Twitter, in which case they’re temporary, scoped, and exactly where they’re supposed to be. Cross-domain script includes? Suspicious, unless they’re loading Google’s reCAPTCHA or your SSO integration, in which case they’re essential.

The report is half of the truth. The contextual analysis is the other half. Without both, you’re making decisions based on incomplete information, and in my experience, those decisions tend to lean toward unnecessary panic and wasted remediation effort.

If you have a VAPT cycle coming up, prepare your stakeholders before the report lands. It’ll save you a week of damage-control conversations that didn’t need to happen.

This article was first published here.

—

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 How to read a VAPT report without panicking appeared first on e27.

The global rush to build artificial intelligence may be making the basic gateway to the internet more expensive for the people who need it most.

That is the uncomfortable warning from the GSMA’s State of Mobile Internet Connectivity Report 2026, released in Hong Kong on September 18. The industry body, which represents mobile operators and the wider mobile ecosystem, says more than 3.4 billion people still do not use mobile internet even though over 90 per cent of them already live within mobile broadband coverage.

In plain terms, the networks are largely there. The devices are not.

Also Read: Policy warning: Without intervention, AI could deepen the digital divide

The report argues that the next digital divide will not simply be between countries with advanced AI models and those without them. It will be between people who can afford the basic tools needed to access AI-enabled services, and those who cannot. For low- and middle-income countries, including several large Southeast Asian markets, that distinction matters. AI in healthcare, education, farming, financial services or public administration means little if the intended users cannot get online in the first place.

“Artificial intelligence has the potential to improve lives on an unprecedented scale, but AI is meaningless if people cannot get online in the first place,” said Vivek Badrinath, Director General of the GSMA.

The smartphone bottleneck

The report’s central point is simple: handset affordability is now the biggest barrier to mobile internet adoption across surveyed low- and middle-income countries, ahead of even digital skills.

By the end of 2025, an entry-level internet-enabled handset cost the poorest 20 per cent of people in low- and middle-income countries the equivalent of 44 per cent of their average monthly income. In Sub-Saharan Africa, that figure rose to 76 per cent.

Those numbers are not abstract. For a low-income worker, a basic smartphone can represent a choice between connectivity and household essentials. In Southeast Asia, where mobile phones are the primary internet device for many users, the cost of entry-level handsets directly affects whether people can access digital payments, government services, online learning, telemedicine and job platforms.

The GSMA says 4.8 billion people now use mobile internet on their own device. But growth is slowing. Around 160 million people came online in 2025, down from 190 million the year before. Meanwhile, 3.1 billion people live within mobile broadband coverage but do not use mobile internet, what the GSMA calls the “usage gap”. Most of them still do not own an internet-enabled device.

This is the gap that policymakers in emerging Asia have struggled with for years. Extending 4G coverage to rural islands, mountain communities or secondary towns is difficult but measurable. Getting affordable devices into people’s hands is harder, especially when household incomes are under pressure and device prices begin to rise.

How AI is pushing up phone costs

The new pressure point is the global AI infrastructure boom. Demand for data centres, servers and AI chips has increased competition for memory and other components also used in smartphones. According to the GSMA report and data from Counterpoint Research, memory prices more than doubled between the third quarter of 2025 and the first quarter of 2026. They then rose by a further 80 to 90 per cent in the second quarter of 2026.

The effect is already showing up in the handset market. Entry-level smartphone prices are rising sharply, while global smartphone shipments are forecast to suffer their largest annual decline on record. The report says the fall is driven mainly by the collapse of the sub-US$100 handset segment, with emerging markets expected to be hit hardest.

Also Read: The digital divide: Islands of modernity in a K-shaped economy

That creates a strange contradiction. AI is being promoted as a tool to widen access to knowledge, improve public services and boost productivity. Yet the same infrastructure race powering AI, one that analysts had already flagged could strain the wider semiconductor supply chain, is pushing up the cost of the devices many people need to use those services.

The GSMA says that until a year ago, reducing the price of entry-level smartphones to US$30 could have made devices affordable for almost 1.6 billion people. Reaching US$20 could have brought affordability within reach for around 2.2 billion people living under mobile broadband coverage. Those price points are now out of reach, the report warns, despite efforts by operators and manufacturers, including the GSMA’s Handset Affordability Coalition.

The body is urging chipset and memory manufacturers to increase the availability of affordable components for entry-level handsets. It is also calling for dialogue among component suppliers, mobile operators, device makers, policymakers and multilateral financial institutions.

That is a broad appeal, but the logic is sound. If AI-driven demand continues to absorb component supply at the high end, the bottom of the smartphone market could be squeezed further. For Southeast Asian countries still trying to close rural, gender and income-based digital gaps, that would be a serious setback.

Why Southeast Asia should pay attention

Southeast Asia is often discussed as a mobile-first region, but that phrase can hide uneven realities. Singapore has near-universal connectivity and is already positioning itself around AI governance, data infrastructure and digital public services. Indonesia, the Philippines, Vietnam, Cambodia, Laos and Myanmar face a more complicated picture, with large populations outside major cities still constrained by affordability, skills or service relevance.

The GSMA’s warning is therefore directly relevant to the region. Many Southeast Asian governments are building digital identity systems, e-payment networks, online tax platforms, telehealth tools and AI-assisted public services. Startups are doing the same in credit scoring, education, logistics, agriculture and small-business software. But these models often assume that the user already owns a capable smartphone and can afford data.

If entry-level devices become more expensive, the business case for inclusive digital services weakens. A farmer cannot use an AI crop advisory tool without a device. A micro-merchant cannot adopt digital bookkeeping without reliable mobile access. A student cannot benefit from personalised learning apps if the household shares one outdated phone.

The economic stakes are significant. Previous GSMA analysis estimates that closing the mobile usage gap would generate US$3.5 trillion in additional GDP between 2023 and 2030, with more than 90 per cent of those benefits flowing to low- and middle-income countries.

Connectivity before AI

The report does not argue that smartphones alone will solve the divide. It also points to digital skills, literacy, online safety, security concerns and the availability of relevant content and services. These issues are familiar across Southeast Asia, where internet access does not always translate into meaningful usage.

But handset affordability is the first gate. Without the device, the rest of the digital economy remains theoretical.

Also Read: The unspoken crisis: Are we building a new digital divide in agriculture?

That makes the GSMA’s report a useful corrective to the current AI conversation. Much of the debate focuses on model capabilities, data centres, regulation and enterprise adoption. Those issues matter. Yet for billions of people, the defining question is more basic: can they afford the phone required to participate?

If governments and the tech industry want AI to be inclusive, they may need to start not with algorithms, but with the low-cost smartphone supply chain. Otherwise, the AI revolution could arrive in emerging markets as another service built for people already connected — while those outside the mobile internet remain exactly where they are.

The post GSMA says smartphone costs could deepen the coming AI divide appeared first on e27.

A founder may want the main business argument. A marketing lead may care about campaign angles. A sales lead may look for customer objections. A busy operator may only want the practical steps. The core idea can be the same, but the way it is framed often needs to change.

That was the problem we kept running into with content.

We would write one solid article, but everyone in the team would find something that they didn’t like.

We were not creating many versions for the sake of having many versions. We needed them because we did not always know which angle would be most useful until we saw the options side by side.

One version might make the article feel relevant to founders. Another might make the same idea clearer for marketers. Another might work better as a sales note. Another might be too weak and get discarded. The purpose of generating several versions was to create a small set of usable choices, not to publish everything.

This mattered because the first version we wrote was often not the best version to distribute. A strong article can still have the wrong opening, wrong example, or wrong emphasis for the audience we want to reach. Multiple versions helped us test the framing before deciding what to post, send, or save for later.

Also Read: The AI economy is quietly exposing what organisations truly value about humans

At first, we used AI the simple way. We pasted the article in and asked it to make a LinkedIn post, an email version, or a shorter summary. That helped, but the results were inconsistent. Sometimes the AI changed too much. Sometimes it kept the wrong parts. Sometimes it made every version sound like a generic marketing post.

The issue was that we were asking AI to rewrite content without giving it the workflow around the rewrite.

So we changed the process and used an agent can be taught to refer to specific predetermined rules and style whenever we feed it a new document.

Instead of asking for random variations, we started by defining the audiences and creating an agent for each. Then we would define what each reader was likely to care about, what they might ignore, and what kind of example would make the article feel relevant to them.

Only after that would the AI create the variants.

That small change made the output more useful.

For a founder, the version might focus on leverage, team capacity, and how repeated work can become a system. For a marketer, the same idea might focus on campaign reuse, content distribution, and message testing. For a sales lead, it might focus on follow-ups, lead context, and not losing useful details between conversations. For an operator, it might focus on handoffs, approvals, recurring checks, and reducing manual follow-through.

This did not mean publishing all four versions back-to-back to the same audience. Most of the time, we used only one. Sometimes we saved another for later. Sometimes one became a LinkedIn post, another became an email intro, and another became a sales note. The value was not volume for its own sake. The value was seeing the possible angles before choosing the strongest one.

The article was not completely different each time. It should not be. The point was not to create unrelated content. The point was to make the same idea easier for different readers to enter.

Also Read: “AI amnesia” is quietly costing Southeast Asian brands their customers

The review step still mattered. AI can adapt framing quickly, but it can also drift. It may exaggerate a claim, add examples that were not in the original, or make the tone too polished. So, we added a simple review rule: every version had to keep the original meaning, avoid adding unsupported claims, and still sound like something a real person in that role might write.

That meant we did not approve every version immediately. We checked whether the opening matched the audience. We checked whether the examples still made sense. We checked whether the conclusion still pointed back to the original idea. If the AI version sounded impressive but no longer matched the article, we rejected it or narrowed it again.

This workflow also helped us avoid making every post sound the same. Without the audience brief, AI tends to produce similar openings: “In today’s fast-paced world,” “AI is changing everything,” or “Businesses need to adapt.” Those lines may be serviceable, but they do not feel specific. When the audience is clear, the opening can become sharper.

A founder version can begin with capacity. A marketer version can begin with distribution. A sales version can begin with lead leakage. An operator version can begin with handoffs. Same article, different doorway.

Before this workflow, different team members might rewrite the same article in different ways and accidentally change the message. With the AI workflow, the source article stays fixed. The audience brief gives direction. The review step catches drift. The final versions can sound different without becoming disconnected.

That matters because content reuse is easy to do badly. You can paste the same article everywhere and bore people. Or you can rewrite it so aggressively that the original point gets lost. A structured AI workflow gives a middle path: adapt the framing, keep the core idea, and review the output before publishing.

—

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 How we use AI to turn one article into audience-specific versions appeared first on e27.

Crunchbase’s first-half 2026 numbers looked like an industry-wide win: global venture funding hit US$510 billion, smashing the previous half-year record. Look closer and the picture curdles. OpenAI and Anthropic alone absorbed 43 per cent of every venture dollar deployed on the planet in six months, with Anthropic’s US$65-billion raise alone equal to nearly a third of the whole quarter’s global funding.

AI’s share of capital jumped from under 50 per cent a year earlier to more than 70 per cent by the second quarter, while deal count barely moved. This was concentration, not a broader boom.

Southeast Asia’s own numbers make the gap concrete. The region raised just US$2.81 billion across 98 equity deals in the first quarter of 2026, the lowest quarterly deal count in at least eight years, with more than 70 per cent of that value coming from a single Series C. Strip that round out and SEA raised roughly US$800 million in three months, a rounding error next to two companies’ first-half haul, more than 40 times the region’s entire 2025 total.

The uncomfortable truth: this isn’t a US problem SEA can watch from a distance. Sovereign capital, including Temasek-linked vehicles, already sits inside the AI mega-rounds reshaping the market, and concentration is becoming how the region’s own capital behaves too.

Read the full story here.

REGIONAL

Grab buys 60 per cent of Atome for US$1.49B in cash, Grab’s biggest fintech acquisition gives it a second underwriting data set beyond ride-hailing, with a formula-priced second tranche that could push the total value to US$4.5 billion by 2028.

Databricks commits US$350M to expand Singapore hub, the AI and data giant will quadruple its Singapore footprint, growing headcount from 250 to over 500 as regional enterprises push AI pilots into production despite thin governance maturity.

SBI Group backs Singapore payments firm DTCPay’s US$25M round: Japanese financial giant SBI Group has joined DTCPay’s Series A round, bringing total funding to US$25M. The Singapore-based crypto-fiat payments firm serves over 2,000 merchants across Southeast Asia.

Waymo picks Singapore for its next robotaxi market, Alphabet’s autonomous-vehicle unit will begin manual test driving in 2027 before opening a commercial robotaxi service in 2028, joining Tokyo, London and Munich in its overseas expansion.

GreenSM launches electric ride-hailing motorbikes in Jakarta: The Indonesian startup entered Jakarta’s competitive ojol market with an all-electric fleet, targeting a cleaner alternative to petrol-powered ride-hailing. No fleet size or funding details were disclosed at launch.

Plaud doubles Singapore bet to US$15.7M, hires 100, the Chinese-founded voice-recorder maker grew its Singapore team tenfold in nine months, though its own revenue figures remain disputed even as rivals Notta and Amazon’s Bee crowd the category.

Maybank sees Atome adding up to US$220M EBITDA by 2028, the bank’s research arm called Grab’s Atome valuation full but not excessive, while flagging that ecosystem synergies, not just direct earnings, are what investors are being asked to underwrite.

Singapore’s non-oil exports jump 46.2 per cent on AI chips, integrated circuits and other AI-linked electronics drove August’s export surge, extending July’s growth and reinforcing the AI hardware cycle as a rare bright spot in trade data.

CXA’s Rosaline Chow Koo returns with broker AI tool Covee, the insurtech veteran’s new US$750,000 pre-seed venture compresses benefits-renewal work from days into minutes for boutique brokers, a narrower bet than her earlier, capital-heavy CXA.

Cambodia’s Newwave lands CIFC backing for regional AI push, the Phnom Penh AI firm, which already earns more than 70 per cent of revenue from abroad, becomes only the second portfolio company for Canadia Group’s young corporate venture fund.

Ricoh backs Singapore’s Intellect in first fund bet, Japan’s office-services giant chose mental health as the theme for its debut cheque from a new US$20-million innovation fund, betting workplace wellbeing data complements its own.

FEATURES AND INTERVIEWS

SEA robotics’ record year hides a one-company problem, a single US$670-million round drove 96 per cent of the region’s 2026 robotics funding, and Singapore alone captured 92 per cent of the total, concentration dressed up as growth.

Why FinSight still has zero Southeast Asian bets, the late-stage fund that demands 35 per cent annual returns in dollars has backed Venezuela and Uzbekistan but not a single SEA company — though it says that is about to change.

INTERNATIONAL

Hong Kong’s CoinEx shuts down after nine years: The crypto exchange cited unsustainable operating conditions as it winds down, marking one of the more significant exchange closures in Asia. Users have been advised to withdraw funds before the deadline.

India imposes 0.4 per cent UPI fee on large payments, the world’s largest real-time payments network will charge merchants on transactions above ₹2,000 from October 15, ending a zero-fee model that has made UPI free to accept since 2020.

Hong Kong’s first five-year plan puts AI, fintech first, the blueprint creates an AI Commissioner post and expands the HKMA’s generative-AI sandbox to insurance and wealth management, alongside fresh venture-fund commitments.

Temasek-backed Xora leads Hang Ten’s US$53M seed round, the enterprise-AI services firm closed its second seed round in five weeks, taking total funding to US$85 million as Aramco Ventures and Mayfield also joined.

How Hong Kong tries to solve tech’s hardest problem, HKSTP’s venture fund has attracted roughly HK$13 in private capital for every HK$1 it invests, part of a broader push to bridge Asian deep-tech research and commercial scale.

China’s Apex Intelligence raises US$50M for self-evolving AI, the Beijing startup, founded by a Tsinghua professor, is building foundation models that refine themselves through iterative experiments for chip design and drug discovery.

Taiwan’s Boba Tech Summit puts Gen Z founders on stage, YouTube and Twitch co-founders shared billing with still-building Gen Z entrepreneurs at Taipei’s inaugural summit, part of the island’s push beyond its semiconductor-supplier image.

A practical resource map for founders entering Taiwan, the guide urges founders to treat visas, incorporation and office leases as separate decisions, sequencing customer validation before committing to any programme or address.

Ex-TikTok staff build an AI app that fixes your poses, Superpose, built by former TikTok product leads, has been downloaded over 22,000 times since July and generates AI-suggested poses from a single selfie.

CYBERSECURITY

Indian police probe Google over 500,000 fake Gmail accounts: Authorities are questioning Google over a mass creation of fake Gmail IDs used to send bomb hoax messages, raising urgent questions about platform accountability and identity verification at scale.

SEMICONDUCTOR

Applied Materials commits US$5B to India chip push, the equipment maker will build a 140-acre research park and expand its supply chain tenfold over a decade, adding to its recent US$500-million Singapore manufacturing expansion.

India targets 200 chip design firms and 100,000 workers: Under its Semicon 2.0 initiative, India is pushing to build a domestic chip design talent pipeline and firm base, signalling a shift from assembly-focused to design-led semiconductor ambitions.

Nexperia and Tata Electronics deepen India chip ties, the Dutch chipmaker will manufacture at Tata’s new Gujarat and Assam facilities, giving India’s fledgling semiconductor ecosystem a global customer before its first fab even ships product.

AI

Manus seeks US$4B valuation in first round post-Meta split: The AI agent startup is raising its first external funding since parting ways with Meta, with the round expected to attract significant investor interest given the firm’s autonomous agent capabilities.

‘AI amnesia’ is quietly costing SEA brands customers, Seven in 10 APAC consumers have abandoned an AI chat mid-conversation over lost context, and five per cent walked away from the brand entirely, Twilio’s new research found.

OpenAI catches its own models hiding mistakes from users, GPT-5.6 Sol left instructions for its successors to conceal errors, a disclosure made under a new misalignment-reporting framework as rival labs race toward IPOs and trillion-dollar valuations.

AI infrastructure spending is far from peaking, says BIMB, global data-centre capex could grow fortyfold to US$31.6 trillion by 2050, the research house said, with no sign yet of the budget cuts or overcapacity that would signal a downturn.

Only some APAC tech firms would survive an AI slowdown, S&P Global Ratings stress-tested chip foundries, memory makers and server assemblers against a slower AI-spending scenario, finding TSMC-style market leaders far more resilient than memory-focused rivals.

THOUGHT LEADERSHIP

Why Singapore investors hold more Apple than Singtel, a study of 82 retail portfolios found no SGX stock cracked the top 30 holdings, a ‘reverse home bias’ the author says AI-assisted investing risks making worse, not better.

The founder-to-minister pivot isn’t ASEAN’s real problem, Nadiem Makarim’s arc from Gojek founder to minister exposes missing governance infrastructure, not a flawed career move, the writer argues, urging founders to set boundaries before office beckons.

Mobile’s next US$4.5 trillion won’t come from connectivity, operators capture barely US$170 billion of the US$3.7 trillion in value mobile creates, and engagement, not spectrum, is the real bottleneck, the telecoms writer argues.

Does the iPhone show us which jobs AI will destroy first?, small, routine transactions disappear before big ones, the way compact cameras vanished once smartphones bundled photography in — a lens the writer applies to translation, transcription and basic design work.

Some Gen Z graduates are skipping AI degrees, and winning, paid corporate apprenticeships now beat a four-year AI degree by roughly US$600,000 over four years, the writer argues, urging Asian founders to hire on demonstrated work over credentials.

Your team finished the AI course. Can it challenge the machine?, course completion measures exposure, not competence, the writer argues, proposing a 20-minute exercise testing whether staff can spot a wrong AI answer.

The AI boom won’t help you scale. Unit economics will, regional VC funding still hasn’t returned to its 2021 peak, the writer argues — what looks like an AI boom is really a sorting event rewarding businesses with real governance maturity.

What Maybank’s US$10B bet reveals about market readiness, Malaysia ranks third globally for expat-friendliness, yet opening a bank account there remains genuinely difficult, the writer argues, using Brazil’s banking infrastructure as an instructive counter-case.

Nobody gives you time to explain. That’s the real problem, investors spend barely two minutes on a first-pass pitch deck, the writer argues, describing a venture built on investment logic and film grammar rather than longer decks.

The creator revenue line Southeast Asia hasn’t priced yet, AI systems cite YouTube roughly 50 times more often than TikTok, leaving the region’s TikTok-heavy creator economy invisible to the assistants now shaping purchase decisions, the writer argues.

Asia’s research-tech firms are big, yet strangely invisible, Singapore’s PatSnap and Engine Biosciences serve millions of scientists but skip the mainstream press that funders actually read, the writer argues, urging founders to pitch research integrity as news.

Bitcoin and Ethereum moved in lockstep after the Fed’s hike, the synchronised rally was a relief bounce, not a fundamental repricing, the columnist argues, pointing to falling open interest and persistent Bitcoin-ETF outflows as signs of fragile momentum.

The capital-driven rise of modern business software suites, Odoo’s rapid growth has left Singapore SMEs frustrated with thin base modules and escalating hidden costs, a sponsored analysis argues, pointing to ERPNext, Zoho One and other alternatives.

The post Ecosystem Roundup: The AI funding boom has a Southeast Asia-sized blind spot appeared first on e27.

For years, stablecoins have carried a simple promise: move money globally with the speed of the internet, without the cost and delays of traditional banking rails. The harder question has always been whether they can move from crypto-native circles into regulated, everyday payments.

Singapore-based dtcpay is betting that the answer lies not in bypassing the financial system, but in building tightly licensed infrastructure that lets stablecoins sit alongside fiat money.

The company has completed its US$25 million Series A round, adding Japan’s SBI Group as a strategic investor after Vertex Ventures Southeast Asia & India led the initial tranche in April 2026.

Also Read: SEA’s stablecoin boom has a dollarisation problem nobody’s pricing in

SBI is investing through SBI Ventures Asset and the SBI-NTU-Kyobo Digital Innovation Fund. The round also includes Genedant Capital and existing investor Kwee Liong Tek, a Singaporean business figure who has continued to back the company.

The fundraise gives dtcpay more capital at a time when stablecoins are drawing renewed attention from banks, payment companies and regulators. Unlike volatile cryptocurrencies such as Bitcoin, stablecoins are digital tokens typically pegged to fiat currencies such as the US dollar. In theory, that makes them more useful for payments and settlement. In practice, adoption still depends on licensing, banking access, merchant acceptance and whether consumers see any reason to use them.

Building payment rails, not just a crypto wallet

Founded by Alice Liu and Band Zhao, dtcpay provides infrastructure that allows businesses and individuals to accept, hold and transact using stablecoins and fiat currencies. Its system includes a real-time swap engine that converts between stablecoins and traditional money, aiming to reduce the operational friction that often comes with handling digital assets.

The company positions itself against the long-standing pain points of cross-border payments. Many international transfers still rely on correspondent banking networks and SWIFT messaging, which can involve multiple intermediaries, opaque fees and settlement times stretching over several days. This is particularly relevant in Southeast Asia, where businesses frequently operate across fragmented currencies, banking systems and regulatory regimes.

dtcpay’s proposition is that stablecoins can help compress settlement time and cost, but only if wrapped inside regulated payment infrastructure that merchants and institutions can trust.

The company has already pushed into several commercial use cases. It launched a Digital Payment Token point-of-sale acceptance product, enabling merchants to accept stablecoin payments in physical stores. It also integrated with WalletConnect, giving it access to more than 700 wallets used by consumers globally.

On the consumer side, dtcpay partnered with Visa to introduce a stablecoin-to-fiat Visa Infinite card for customers in the region. The card allows users to spend across fiat and stablecoin balances at more than 150 million merchant locations worldwide, according to the company.

Also Read: Southeast Asia can’t simply license its way to stablecoin sovereignty

In Singapore, dtcpay has also worked with BNB Chain on stablecoin adoption and enabled department store Metro to accept stablecoin payments. Hospitality partners such as Capella Singapore have also been part of its early merchant network.

These examples matter because stablecoin payments have often struggled to break out of online trading and treasury use cases. For adoption to deepen in Southeast Asia, the technology needs to work in settings that are familiar to both merchants and consumers: retail checkouts, corporate payments, travel, remittances and cross-border trade.

Regulation as a growth strategy

dtcpay’s biggest selling point is not simply its technology, but its licensing posture. The company is a Major Payment Institution licensed by the Monetary Authority of Singapore. It also holds an Electronic Money Institution licence in Luxembourg, allowing it to provide regulated payment services across the European Economic Area. In addition, dtcpay says it holds licences and registrations in Hong Kong, Australia, the US and Canada.

That regulatory footprint gives dtcpay a base from which to pursue both Asian and Western markets. It also reflects a broader shift in digital assets: after years of offshore experimentation, institutional capital is now gravitating towards companies that can meet compliance requirements in major financial centres.

Singapore has been central to that shift. The city-state has tightened rules around crypto speculation while continuing to support regulated digital asset infrastructure, tokenisation and cross-border payment experiments. For startups such as dtcpay, that creates both an opportunity and a constraint. The market rewards regulatory discipline, but moving too slowly can allow global competitors to capture corridors before regional players scale.

The fresh funds will be used to expand dtcpay’s product suite and merchant network, as well as support its product roadmap for the rest of 2026. The company plans to launch a revamped business portal for enterprise clients and add more consumer-facing features to the dtcpay app.

“We did not raise this round to sustain what we have built. We raised it to fundamentally change how money moves across borders,” said Liu, founder and CEO of dtcpay. She added that SBI’s backing validates the view that compliant, real-world stablecoin payments are “not a distant vision but an infrastructure being built right now.”

Why SBI’s entry matters

SBI’s participation gives dtcpay more than a financial investor. The Japanese group operates across securities, banking, insurance, asset management and digital assets, and has long been active in fintech and blockchain-related infrastructure. Its Singapore arm, SBI Ven Capital, manages the SBI-NTU-Kyobo Digital Innovation Fund, which was launched in 2022 to invest in early-stage digital transformation and platform companies across Southeast Asia.

For dtcpay, SBI could help open doors in Japan and across institutional financial networks. For SBI, the investment fits a wider regional strategy as Japanese financial groups look beyond a mature domestic market and seek exposure to Southeast Asia’s faster-growing digital economy.

“dtcpay has made decisive progress in establishing itself as the region’s leading regulated payment infrastructure that bridges traditional payments and stablecoins,” said Eiichiro So, CEO of SBI Ven Capital. He added that the investment marks the start of a strategic partnership and supports SBI’s aim to expand digital asset corridors between Japan and Southeast Asia.

Genedant Capital, a Singapore-based fund manager with more than US$2 billion in assets under management and advisory, brings a different network of family offices, private wealth investors and institutional relationships. Vertex, meanwhile, gives dtcpay access to a venture platform with a long history of backing Southeast Asian technology companies.

A crowded race for stablecoin payments

dtcpay is not alone in trying to make stablecoins usable for mainstream commerce and financial institutions. In Singapore, Triple-A has built crypto payment acceptance infrastructure for merchants, while StraitsX, part of Fazz, focuses on regulated stablecoin issuance and digital payment infrastructure.

Globally, companies such as Circle, Ripple and BVNK are pursuing various parts of the same market, from stablecoin settlement and treasury tools to cross-border payment rails for businesses.

Also Read: Nium acquires Cypher as fiat and stablecoin payments converge

The competitive question is whether dtcpay can convert its licences, merchant integrations and investor network into scale. Stablecoin payments are still early, and many users remain indifferent to what rails sit underneath a transaction as long as it is fast, cheap and reliable. That means dtcpay’s success will depend less on convincing the public to “use stablecoins” and more on making the experience feel no different from existing digital payments.

That is a difficult but potentially large opening. Southeast Asia’s businesses already operate across borders, currencies and platforms. If regulated stablecoin infrastructure can reduce settlement delays without adding compliance risk, it could become a practical layer in the region’s payment stack.

dtcpay’s US$25 million round suggests investors believe that moment is getting closer. The harder work now is proving that stablecoins can become not just a financial market instrument, but a routine way for people and companies to move value.

The post SBI joins dtcpay’s US$25M round to bridge Japan, SEA stablecoin corridors appeared first on e27.