Day: August 23, 2026

For years, Southeast Asia’s startup economy has rewarded speed. Founders were expected to launch quickly, localise faster than global rivals, and chase market share across a region where digital adoption often outpaced regulation. That instinct still matters. But as artificial intelligence moves from pilot projects into banking, healthcare, government services, logistics and public infrastructure, speed is no longer enough.

A different test is emerging: trust.

Malaysia’s National AI Action Plan 2026-2030, or AI Nation 2030, makes this shift explicit. One of its core foundations is “Trust via Responsible Governance”, a signal that the country wants AI adoption to be measured not only by productivity gains or startup growth, but also by whether systems can be explained, audited and governed.

Also Read: Malaysia’s AI Nation 2030 puts cities and farms at the heart of climate resilience

For startups, this should not be read simply as another compliance burden. In the coming corporate AI market, governance may become a sales advantage. Enterprise buyers will increasingly ask not just what an AI product can do, but how it was built, what data it uses, where the risks sit, and whether those risks can be defended in front of boards, regulators and shareholders.

In other words, the next competitive edge in AI may belong to companies that can make trust operational.

From black-box tools to board-level accountability

The clearest sign of this shift is the AI-Aware Stewardship initiative under Malaysia’s AI Action Plan. The policy targets corporate boards and seeks to prepare them to oversee risks linked to AI and emerging technologies.

The targets are specific. Malaysia wants 30 per cent of large public listed companies to adopt emerging technology governance best practices by 2028, rising to 50 per cent by 2030. To move this from aspiration to practice, the Securities Commission Malaysia, Bursa Malaysia, the National AI Office, the Personal Data Protection Commission, and the Ministry of Science, Technology and Innovation are coordinating updates to the Malaysian Code on Corporate Governance and Listing Rules.

Large listed companies are expected to be encouraged to publish an Emerging Technology Governance Statement in their annual reports. They may also use maturity scorecards to show shareholders how prepared they are to manage digital and AI-related risks.

That changes the buying environment for B2B startups. A bank, telco, insurer or major retailer will find it harder to adopt a black-box AI product if it cannot explain how the system works, what safeguards are in place, or where accountability lies when something goes wrong. Procurement teams may still care about price and performance, but boards will increasingly care about audit trails.

For founders selling into large enterprises, this means the product demo is no longer enough. The due diligence file matters just as much.

The rise of audit-ready AI

One of the more demanding elements of the new framework is the push for risk mapping. Large companies will be encouraged to maintain a board-approved AI system inventory and risk assessment map, subject to internal audit review.

Also Read: Malaysia wants 300,000 AI jobs by 2030. Talent will decide if it gets there

This has direct implications for vendors. If a startup’s product sits inside a corporate AI inventory, the enterprise customer will need details about the system’s model, data, dependencies, controls and failure risks. A vendor that cannot provide these details may slow down the buyer’s approval process, or be dropped altogether.

The practical response is for startups to become audit-ready by design.

That starts with data provenance. Founders need to know where their training and operational datasets came from, how personally identifiable information was handled, whether data was licensed properly, and how usage aligns with Malaysia’s emerging data-sharing frameworks, including the Akta Perkongsian Data 2025.

It also requires model explainability. Not every AI system can be made simple, especially those built on complex machine learning methods, but startups should be able to explain how decisions are generated, what variables matter, and where human oversight is required.

Bias and safety logs will also become more important. Startups should be able to show how they test for unfair outcomes, handle edge cases, document incidents, and update models when risks appear. This is especially relevant in Southeast Asia, where AI tools often operate across multiple languages, dialects, income groups and cultural contexts. A model trained for one market may behave differently in another.

The companies that build this documentation early will have an advantage. They can plug more easily into enterprise governance processes, shorten procurement cycles, and reassure investors that the business will not collapse under regulatory scrutiny as it scales.

A risk-based route for founders

A common fear among startups is that AI regulation will favour incumbents with large legal teams. Malaysia’s plan appears to recognise this risk by proposing a hybrid, risk-based governance framework.

Under this approach, not all AI systems are treated the same. Lower-risk applications can operate under voluntary guidance, while higher-stakes uses in areas such as finance, healthcare or communications may face tighter rules overseen by sector regulators, such as Bank Negara Malaysia or the Malaysian Communications and Multimedia Commission.

This distinction matters. A startup building an AI tool for internal workflow automation should not face the same burden as one automating credit decisions or clinical recommendations. Risk-based governance, if implemented clearly, can give young companies room to innovate while giving enterprises a clearer path for adoption in sensitive sectors.

Also Read: From paddy fields to small shops, Malaysia maps an inclusive AI future

Malaysia is also introducing a National AI Classification initiative, led by the National AI Office, to certify “Made-by-Malaysia” AI systems. The certification is expected to evaluate the AI lifecycle, from compute and data layers to the final model.

For local startups, this could become more than a badge. Certified companies may gain prioritised access to the National Data Exchange, compute voucher programmes, local supply chain registries, government procurement opportunities and large corporate tenders. That would make governance a market access tool, not just a legal exercise.

Why this matters beyond Malaysia

Malaysia’s approach also sits within a broader Southeast Asian moment. Governments across the region are trying to balance AI adoption with public trust. Singapore has pushed governance through tools such as AI Verify, Indonesia and Thailand are examining digital rules through their own policy lenses, and ASEAN has been building regional guidance for responsible AI.

For startups, the regional lesson is simple: compliance designed only for one buyer or one jurisdiction will not be enough. A Malaysian startup seeking to sell across ASEAN should design internal controls that can travel. That means aligning safety, data and documentation practices with international standards and emerging regional frameworks, including the ASEAN AI Safety Network.

This is particularly important because Southeast Asian startups often scale regionally before they are fully mature internally. A company may start with a Malaysian bank, then pitch a Singaporean insurer, an Indonesian fintech, or a Philippine conglomerate. Each buyer may have different rules, but all will increasingly ask similar questions about data, accountability and risk.

Governance as a growth engine

For founders, the roadmap is becoming clearer. Start with an internal AI register that maps models, datasets, third-party APIs, security controls and human oversight points. Train engineering, product and leadership teams to understand responsible AI, not as a slogan but as part of product management. Build documentation that can withstand review by enterprise risk teams, investors and regulators.

The bigger point is cultural. AI governance should not sit only with lawyers at the end of a sales process. It needs to be built into product design, model development, customer onboarding and post-deployment monitoring.

Also Read: Malaysia’s sovereign AI bet: Local context becomes the next startup moat

Malaysia’s AI Nation 2030 plan suggests that the region’s AI market is entering a more mature phase. Startups that treat governance as paperwork may struggle. Those that treat it as infrastructure may find it opens doors.

The next wave of AI adoption in Southeast Asia will not be won by the fastest builders alone. It will be won by companies that can show their systems work, explain why they can be trusted, and prove they are ready for the scrutiny that comes with scale.

The post Why Malaysia’s AI Nation 2030 plan matters for B2B startups appeared first on e27.

Three months ago, I sat in a quarterly risk committee meeting at an Indonesian bank, watching a climate risk update presented in twelve slides over fifteen minutes. The presentation covered taxonomy alignment, sustainable finance commitments, and progress against the bank’s net-zero pathway. It was professional, well-researched, and accurate. It also did not mention the bank’s exposure to physical flood risk across its real estate book, the transition risk inside its coal-related loans, or the basis on which any of those risks were being priced into provisioning.

After fifteen years inside Indonesian risk functions, I have come to see that pattern as the defining shape of climate risk inside the region’s banking sector. The reporting infrastructure has matured rapidly. The provisioning infrastructure underneath has not. The gap between what banks disclose about climate and what their balance sheets actually carry has become the most consequential unpriced exposure in ASEAN banking.

The framework that was built

Indonesia’s OJK has, over the past three years, built one of the more thoughtful climate risk frameworks in ASEAN. The Sustainable Finance Roadmap, the green taxonomy, the climate disclosure requirements, the architecture is in place, and Singapore, Malaysia, and the Philippines have moved in parallel. Most major banks now publish annual climate disclosures, often aligned to TCFD recommendations. The disclosures are real work. They are not the same thing as risk management.

Where the exposure actually sits

Three categories of climate exposure inside Indonesian bank balance sheets are visible enough to name and large enough to matter.

Physical climate risk in property and infrastructure. A significant share of commercial real estate financing sits in coastal cities exposed to subsidence, tidal flooding, and increasingly severe wet-season rainfall. The collateral underlying these loans is rarely revalued against forward-looking climate scenarios. The provisioning logic assumes the asset retains its current value. The asset, increasingly, does not.

Transition risk in carbon-intensive sectors. Loans extended to coal, palm oil, and heavy industrial sectors carry exposure to a rapidly evolving regulatory environment, domestic carbon pricing, the European Union’s deforestation regulation, and sector-specific phase-out commitments. The credit framework that priced these loans five years ago did not anticipate that some underlying assets could become stranded inside the loan tenor.

Also Read: Indonesia’s AI hiring gap is real, just not 28×

Cascading climate risk in adjacent sectors. The most under-discussed exposure is not the direct one. It is the credit risk inside borrowers whose own portfolios, supply chains, or customer bases are climate-exposed. A logistics company is not a climate-exposed borrower in the conventional sense. A logistics company whose largest customer is a flood-prone factory is.

Why the framework misses it

The disclosure architecture and the provisioning architecture were built for different purposes, and they have not been reconciled.

Disclosure frameworks make the institution’s climate position legible to external stakeholders. They are not designed to drive loan-level loss provisioning, capital adequacy, or pricing inside the bank.

Provisioning frameworks were built before climate was on the regulatory radar. The expected credit loss model accommodates forward-looking information in principle, but most banks still apply it with historical loss data and short-horizon scenarios. Climate risk operates on a longer horizon than the provisioning logic was built for.

What is starting to work

A few institutions are beginning to close the gap.

Climate-adjusted credit reviews. Some banks now incorporate climate scenarios into credit committee processes for large or long-dated exposures. The discipline of forcing the question into the same room as the lending decision is producing more honest pricing.

Sector concentration limits with climate triggers. Some institutions set internal limits on exposure to high transition-risk sectors and lower those limits as policy clarity improves. The mechanism is imperfect. It is the closest thing to a working transition risk control I have seen in the region.

Also Read: How to get beyond the chatbot and boost your AI productivity

Collateral revaluation under climate scenarios. The most rigorous response I have seen comes from institutions revaluing real estate collateral under multiple climate trajectories, not just the central case. The revaluation rarely changes a single loan’s status. It consistently changes the capital the bank holds against the portfolio.

What needs to happen

Three moves would meaningfully reduce systemic exposure.

Connect disclosure to provisioning. The climate analyses that flow into TCFD-style reports should also flow into expected credit loss calculations, capital planning, and pricing. The reports and the reserves should be telling the same story.

Require forward-looking collateral valuation for long-dated exposures. Where loan tenors extend across plausible climate horizons, the collateral assumption should be tested against those horizons rather than against present-day comparables.

Bring transition risk into supervisory stress testing. ASEAN supervisors already run credit, liquidity, and market stress tests. They should be running transition stress tests, modelling specific policy scenarios across carbon-intensive sectors and measuring portfolio capital impact.

The macro stakes

Indonesia is among the most climate-exposed major economies in the world, with a banking sector whose stability matters regionally. The disclosure architecture the country has built is genuinely good. The provisioning architecture has not caught up.

The climate risk inside Indonesian bank portfolios is not theoretical. It sits on balance sheets now, accruing exposure that is not being priced, against scenarios the institutions’ own disclosures already say are coming. The bill, when it arrives, will not be paid by the disclosure framework. It will be paid by the capital base. The window to close that gap is closing.

—

The post Climate risk’s invisible threat: What ASEAN banks aren’t accounting for appeared first on e27.

Singapore’s fight against scams is moving deeper into the messaging apps people use every day, after police disrupted more than 30,000 Apple iMessage accounts linked to a campaign that has already caused about SGD2.2 million (US$1.7 million) in losses.

The Singapore Police Force said its Cyber Command has been detecting and disabling accounts tied to the scam since June 2026. The loss figure has climbed quickly: on August 5, police had put the damage at SGD1.2 million (~US$940,000). In other words, reported losses rose by nearly US$800,000 in a matter of weeks.

Also Read: Almost got “digitally arrested.” India needs Singapore’s playbook before the next scam call

The case underlines a problem that is becoming familiar across Southeast Asia: scammers are not relying only on old-fashioned SMS blasts or suspicious phone calls. They are moving across encrypted messaging apps, social platforms, marketplaces and ad networks, looking for whichever channel has the least friction and the most trust.

In this campaign, fraudsters sent iMessages pretending to be courier companies such as NinjaVan, J&T Express and SPX Express, as well as government agencies and financial institutions. The messages directed recipients to spoofed websites designed to look like the real thing. Victims were then asked to make a small payment or settle a fine, often by entering card or banking details.

The amounts requested may have seemed minor, but the information handed over was valuable. According to police, some victims who entered one-time passwords later discovered that their cards had been added to mobile wallets, bank security tokens had been registered on unfamiliar devices, or their accounts had been accessed without permission. Many only realised what had happened after seeing unauthorised transactions.

Why iMessage is a harder target

The campaign also exposes a regulatory and technical gap. In Singapore, SMS scams have been targeted through network-level filters and a sender ID registry, which helps prevent fraudsters from impersonating trusted organisations through text message headers.

Also Read: Phishing threats: Protecting your online shopping and banking

But iMessage runs on Apple’s own system, outside the traditional telecoms layer. That means it is not covered by the same filters and registry used for SMS. For a scammer, that difference matters. A message delivered in Apple’s blue bubble can appear familiar and personal, especially to users who do not think of iMessage as a risky channel.

Police stressed that government agencies and courier companies do not use iMessage to communicate with the public. That simple point is important because many delivery-related scams rely on timing and plausibility. In a city where online shopping, food delivery and parcel tracking are part of daily life, a message about a failed delivery or unpaid fee can feel routine enough to click.

Singapore is not alone in facing this shift. Across Southeast Asia, fraud groups have become more sophisticated in blending social engineering with real consumer habits. Delivery scams, fake toll or tax notices, investment fraud and phishing links often travel through the same apps people use to speak with family, sellers, banks and colleagues. The more commerce moves into chat, the more attractive these channels become.

New codes put pressure on platforms

The iMessage disruption comes shortly after Singapore issued new Codes of Practice under the Online Criminal Harms Act. Announced on August 17, the codes apply to seven services assessed as posing the highest scam risk: WhatsApp, Telegram, WeChat, Apple iMessage, Apple FaceTime, Google Message and Google Meet.

The services must comply by January 31, 2027, with anti-impersonation measures due earlier, by September 30, 2026.

Also Read: Inside the dark economy of crypto scams: 2024’s most lucrative fraud tactics

Messaging platforms are a major part of the scam landscape. Police said services such as WhatsApp and Telegram accounted for about 23 per cent of scam cases in 2025. That figure is significant because messaging apps are no longer just communications tools. They are customer service channels, sales channels, community spaces and, increasingly, the first point of contact between businesses and users.

For regulators, the challenge is to impose safeguards without breaking the usefulness of these platforms. Identity checks, faster takedowns and impersonation controls may help, but scammers adapt quickly. If one route becomes harder, they often move to another, whether that is an ad, a marketplace listing, a fake account or a compromised device.

This is why Singapore’s approach is widening beyond a single channel. Earlier in the week, police announced a separate Social Media Code covering Facebook, Instagram and TikTok. Together, the three platforms accounted for about 30 per cent of scam cases in 2025, with Facebook alone making up about 18 per cent.

The Social Media Code focuses on scam advertisements, a common gateway for fraud. Platforms will be required to block and promptly remove suspected scam ads, verify advertisers’ identities against government records, and prevent advertisements offering financial products or services unless the advertiser is licensed by the Monetary Authority of Singapore.

That last requirement is particularly relevant in a region where fake investment schemes remain a persistent threat. Scammers often use paid ads to create the impression of legitimacy, sometimes borrowing the faces of public figures, media brands or financial institutions. By the time an ad is reported and removed, victims may already have been funnelled into private chats or fraudulent websites.

Marketplaces also under scrutiny

Singapore is also tightening rules for e-commerce platforms. An enhanced E-Commerce Code covering Carousell, Facebook Marketplace and Facebook Business Pages will introduce stronger controls on logins from unrecognised devices.

Marketplaces have long been vulnerable because they combine informal peer-to-peer transactions with a high volume of listings. Scams can range from fake concert tickets and rental listings to non-delivery of goods and phishing links disguised as payment or delivery pages. Stronger login controls may help limit account takeovers, where criminals use legitimate-looking profiles to trick buyers or sellers.

Also Read: AI phishing is turning trust into APAC cybersecurity’s weakest link

Police said scam cases on services already covered by earlier codes fell by about 37 per cent between 2024 and 2025. That suggests platform rules can have an impact, although the latest iMessage case also shows that fraudsters keep searching for gaps.

The stakes are set to rise further. The government has proposed increasing the maximum penalty for non-compliance to S$10 million (US$7.8 million) per breach. More details are expected when the Scams (Countermeasures) and Other Matters Bill is debated in Parliament in September.

For startups and digital platforms in Southeast Asia, Singapore’s direction of travel is worth watching. The city-state often acts as a regulatory reference point for the region, especially in fintech, digital identity and online safety. Measures introduced there can influence how other markets think about platform responsibility.

For consumers, however, the immediate lesson is more basic: the channel does not guarantee the sender. A message arriving through iMessage, WhatsApp, Telegram, Facebook or TikTok may still lead to the same spoofed payment page. In the current scam economy, trust is no longer attached to the app. It has to be earned at every click.

The post Singapore disrupts 30,000 iMessage accounts as scam losses hit US$1.7M appeared first on e27.

The post US$73,000 and still climbing: How long can Bitcoin ignore the macro storm? appeared first on e27.

J&T Global Express has delivered the kind of first-half numbers that usually make public-market investors sit up.

The Hong Kong-listed logistics company reported revenue of US$7.67 billion for the first half of 2026, up 39.5 per cent year-on-year, while express delivery revenue rose 39.6 per cent to US$7.46 billion. Adjusted net profit more than doubled to US$350.6 million, and adjusted EBIT climbed 121.7 per cent to US$433.6 million.

The headline story is clear enough: J&T is still expanding quickly, still riding the growth of e-commerce, and now important enough to have joined the Hang Seng Index in June. For Southeast Asia, where the company has held the largest express delivery market share for six consecutive years, its performance also speaks to the region’s growing weight inside global logistics.

Also Read: More parcels, less profit: Logistics’ big squeeze

But beneath the strong top-line figures, J&T’s interim results tell a more complicated story. Its fastest growth is no longer coming from China, the company’s largest market by parcel volume. Its improvement in revenue per parcel appears to be driven partly by a shift in geographic mix rather than clear pricing power. And its enlarged HK$2 billion (~US$255 million) share repurchase programme raises questions about how the company is balancing investor returns against long-term network investment.

China is still huge, but growth is slowing

J&T handled 17.50 billion parcels globally in the first half, up 25.1 per cent from a year earlier. Yet that growth was far from evenly spread.

China remained the group’s biggest market by some distance, contributing 11.62 billion parcels, or about two-thirds of total volume. But parcel volume in China grew only 9.6 per cent year-on-year. That is modest compared with the company’s performance elsewhere: Southeast Asia parcel volume jumped 71.2 per cent to 5.52 billion, while “other markets”, mainly Latin America and the Middle East, rose 119.9 per cent to 365 million parcels.

The gap matters because China has long been the world’s most competitive express delivery market. Years of price wars among players such as SF Express, ZTO Express, YTO Express and others have pushed down delivery tariffs and made scale essential. J&T’s market share in China did edge up by 0.5 percentage points to 11.6 per cent, but the single-digit volume growth suggests the company may be bumping into a tougher ceiling in its largest market.

Management has framed this as a move towards better-quality growth. Group Vice President Charles Hou said J&T remains focused on “strengthening operating quality and efficiency”. That is a reasonable priority in a low-margin business. Still, for investors and regional operators, the question is whether China is becoming a cash-heavy but slower-growth base while Southeast Asia and newer markets are asked to carry the expansion story.

The revenue-per-parcel question

One of the more striking parts of J&T’s results is that revenue grew much faster than parcel volume. Overall revenue rose 39.5 per cent, while parcel volume increased 25.1 per cent. On a simple calculation, the company’s average revenue per parcel increased from about US$0.393 in the first half of 2025 to US$0.438 in the first half of 2026.

At first glance, that looks like stronger pricing power. For a logistics company, being able to earn more per parcel while still growing volume is a strong signal. But in J&T’s case, the explanation may be more about geography.

In the first half of 2025, China accounted for roughly three-quarters of J&T’s parcels. By the first half of 2026, its share had fallen to 66.37 per cent. Southeast Asia’s share, meanwhile, rose from 23.05 per cent to 31.54 per cent. Because delivery rates in Southeast Asia and other emerging markets are generally higher than in China’s fiercely competitive domestic market, a larger share of non-China parcels can lift group average revenue per parcel even without a major pricing breakthrough.

Also Read: The rise of logistics startups in Southeast Asia: How AI powers supply-chain revolution

This does not make the improvement meaningless. A healthier geographic mix can support margins, and Southeast Asia’s e-commerce market still has room to grow as online shopping penetrates smaller cities and cross-border sellers seek faster fulfilment. But it does mean the ARPU gain should be read with care. If Southeast Asian markets become more crowded, or if platform-owned logistics arms intensify competition, J&T may face the same pressure on delivery fees that has shaped China’s market.

Southeast Asia is the prize and the battleground

J&T’s Southeast Asian performance remains its strongest argument. The region delivered 5.52 billion parcels in the first half, helped by rising e-commerce adoption, social commerce, and the demand for low-cost delivery across archipelagic and emerging markets such as Indonesia, the Philippines and Vietnam.

The company has also built a dense regional network, including 127 sorting centres in Southeast Asia. That infrastructure is hard to replicate quickly and gives J&T an advantage in markets where delivery reliability can determine whether consumers continue buying online.

But it is not alone. In Southeast Asia, J&T competes with Ninja Van, Flash Express, SPX Express, Lazada Logistics, DHL eCommerce and country-specific postal and courier players. Some rivals are backed by major e-commerce platforms, giving them captive parcel flows. Others are pushing aggressively into small merchants and cash-on-delivery-heavy markets. Globally, J&T’s expansion into Latin America and the Middle East also puts it closer to established logistics groups and regional specialists with deep local networks.

That competitive backdrop makes capital allocation especially important.

A large buyback at a sensitive moment

J&T said it had completed the repurchase of 99.32 million shares and increased the size of its share repurchase plan to US$256.4 million. The company also reported total cash resources of US$2.91 billion, giving it financial room to manoeuvre.

Buybacks are not inherently problematic. They can signal management confidence, improve earnings per share, and return excess cash to shareholders. But for a logistics company still expanding across multiple regions, a repurchase plan of this size deserves scrutiny. The US$256.4 million programme is equivalent to about 73 per cent of the company’s adjusted net profit for the half-year.

The timing is also notable. J&T’s inclusion in the Hang Seng Index brings greater visibility, but also greater pressure from institutional investors and index-tracking funds. A buyback can help support market confidence during that transition. The trade-off is that every dollar used to repurchase shares is a dollar not used to strengthen sorting centres, last-mile capacity, automation, fleet efficiency, or market entry in expensive new geographies.

There is another layer to the numbers. J&T’s announcement highlights adjusted net profit, adjusted EBIT and adjusted EBIT per parcel, but does not foreground statutory net income in the same way. Adjusted metrics are useful for understanding operating performance, especially in businesses affected by non-cash charges or one-off items. Still, investors need the unadjusted picture too, because costs excluded from adjusted earnings can remain economically real.

J&T also promoted a milestone in the second quarter: global average daily parcel volume exceeded 100 million for the first time. That is operationally significant. Yet across the full first half, 17.50 billion parcels over 181 days works out to about 96.7 million parcels a day. The company is clearly operating at immense scale, but the distinction shows how selective framing can make performance appear cleaner than it is.

Also Read: Lazada unveils US$100M affiliate push to power creator-led growth in SEA

For Southeast Asia, the lesson is not that J&T is weakening. It remains a formidable logistics player with regional scale that few competitors can match. The more important point is that its future growth story now depends heavily on this region continuing to expand profitably.

If China keeps slowing and Southeast Asia absorbs more of the growth burden, J&T will have to prove that its regional dominance can translate into durable margins — not just higher group averages created by geographic mix. Its first-half results are impressive. They are also a reminder that in logistics, scale is only half the story. The harder test is whether that scale keeps producing real profits once the easy volume growth fades.

The post J&T Express leans on Southeast Asia as China parcel growth cools appeared first on e27.