Behind the glow of a Telegram chat window, a victim lured by a fake job offer and trafficked into a Southeast Asian scam compound is coerced under violence, running romance scams against strangers half a world away. The wallet collecting the proceeds lies on a public blockchain, and this is just one of the many Southeast Asia crypto scams happening today. This is also one of the uncomfortable intersections Chainalysis maps out in its 2026 Crypto Crime Report. Its data indicates that cryptocurrency flows to suspected human trafficking services reached hundreds of millions of dollars in 2025, an 85% YoY growth. The growth is tracked to the expansion of Southeast Asia’s illicit ecosystem. It’s the very region where scam compounds, online casinos and gambling sites, and Chinese-language money-laundering and guarantee networks, operating primarily over Telegram, feed off each other in an accelerating regional underworld with global reach. A Human-Trafficking/Crypto Trade That Runs Like a Business Chainalysis tracked four categories of suspected cryptocurrency-facilitated trafficking: Telegram-based “international escort” services suspected of trafficking in people, “labour placement” agents that facilitate kidnapping and forced labour for scam compounds, suspected exploitative prostitution networks, and vendors of child sexual abuse material (CSAM). Source: Chainalysis Escort and Prostitution Networks Run Almost Entirely on Stablecoins Nearly half (48.8%) of transfers linked to “international escort” services exceeded US$10,000, a concentration the report says points to organised criminal enterprises operating at scale. “International escort” services and prostitution networks operated on an almost exclusive level via stablecoins. This seemed to suggest that they emphasised payment stability and ease of conversion over the risks of these assets being frozen by centralised issuers. Services were found to be closely tied to Chinese-language money laundering networks, which rapidly enable the conversion of USD stablecoins into local currencies. In doing so, these entities potentially blunt the risk that assets held in stablecoins might be frozen, the report indicated. Scam Compound Recruitment Leaves a Traceable Trail “Labour placement” agent scam operations, especially pig butchering schemes, are deeply intertwined with human trafficking. Victims are lured with fraudulent job offers, then trafficked to scam compounds across Southeast Asia, where they are forced to run romance and investment scams under the threat of violence. Blockchain analysis reveals that recruitment payments typically fall between $1,000 and $10,000, consistent with advertised pricing tiers. This creates identifiable transaction patterns that can be used to detect suspicious activity at scale. These agents also spread their presence across multiple guarantee platforms to maximise reach, with some operating through mainstream cryptocurrency exchanges. CSAM Vendors Turn to Monero and Instant Exchangers Child sexual abuse material vendors, meanwhile, had a tendency to collect payments in mainstream cryptocurrencies. Chainalysis observed that they started using Monero more to launder their proceeds. Instant exchangers, services that allow fast, anonymous crypto swaps with no KYC checks, became a key tool in this process. SEA Trafficking Networks Go Global on Cryptocurrency In 2025, mapping where “international escort” services operate shows that Southeast Asian services, particularly those run in the Chinese language, have expanded worldwide by using cryptocurrency. Based on Chainalysis’ data, Chinese-language services operating across mainland China, Hong Kong, Taiwan, and several Southeast Asian countries have built advanced payment systems and a wide international presence. Source: Chainalysis Large-scale cryptocurrency transactions come in from countries like Brazil, the United States, the United Kingdom, Spain, and Australia. The wide range of countries involved suggests these networks have built the infrastructure needed to operate on a global scale.” As blockchain technology is transparent by nature, this makes it a useful tool for detecting and stopping these activities. Compliance teams and law enforcement can watch for certain warning signs for Southeast Asia crypto scams, such as high-volume transactions through guarantee platforms, wallet clusters linked to multiple types of illicit services, recurring patterns of converting funds to stablecoins, and connections to Telegram channels used for recruitment. Featured image edited by Fintech News Singapore based on an image by kues1 on Magnific The post Crypto Flows to Suspected Human Trafficking Surge 85% YoY, Chainalysis finds appeared first on Fintech Singapore.

Pine Labs has introduced a payment protocol that lets AI agents complete UPI transactions when pre-set conditions are met. The Pine Labs Payment Protocol, or P3P, is now live and is designed for agentic commerce, where AI agents can act on instructions set by users. Current UPI payments still require users to approve transactions at checkout, which can interrupt agent-led purchases. With P3P, a consumer approves a UPI mandate upfront. An AI agent can then complete a payment later if the transaction stays within the approved limits and conditions. This could be used for purchases where timing matters, such as buying digital gold when prices fall or securing a product when it reaches a target price. Pine Labs said users remain in control of the mandate and can update or revoke it at any time. The protocol also includes controls for agent identity, spending limits and audit trails. Gullak, a digital gold savings platform in India, is live on P3P. A Gullak user can set a rule to buy ₹500 of gold if the price falls below ₹16,000 per gram. Once the user approves the mandate, the AI agent can complete the purchase when the condition is met. Pine Labs is also working with other companies across retail, fintech and travel. Vijay Sales, an Indian electronics retail chain with more than 150 stores, is in an active proof of concept with P3P. The protocol could allow customers to set price-based buying rules for items such as smartphones or home appliances. Amrish Rau Pine Labs CEO Amrish Rau said, “In India, UPI’s mandate framework was already architected for agentic commerce. P3P is that layer. An agent securing a flash sale the moment it goes live. A down payment locked in before inventory disappears. A savings trigger at the right price. These are new behaviours, native to how India transacts. Pine Labs is building the commerce infrastructure for that world.” P3P is currently live on UPI. Pine Labs is also working with major card networks to extend the protocol to card transactions.     Featured image: Edited by Fintech News Singapore, based on image by Design-Marjolein via Magnific   The post Pine Labs Brings AI Agent Payments to India’s UPI appeared first on Fintech Singapore.

Bizcap Singapore has launched the Bizcap Frequent Funders (BFF) Program, a new partner loyalty initiative designed to reward high-performing brokers and referral partners with exclusive incentives and luxury experiences as they achieve funding milestones. The launch follows Bizcap’s successful first year in Singapore, where the non-bank lender recently celebrated its one-year anniversary and increased its lending limit to S$1 million, enabling it to support businesses with larger funding requirements The BFF program has been created to recognise and reward partners who consistently help Singapore SMEs access the capital they need to grow. Through a tiered rewards structure, brokers and partners who reach quarterly funding milestones will unlock premium experiences and exclusive rewards. Accredited partners are automatically enrolled in the program, with no registration required. The program includes three reward tiers: Bronze status: Fund $300,000 or more within a quarterly period and receive a reward valued at $888 Silver status: Fund $600,000 or more within a quarterly period and receive a reward valued at $1,888 Gold status: Fund $1 million or more within a quarterly period and receive a reward valued at $3,888 Joseph Lim, Bizcap’s Managing Partner for Asia, said the initiative reinforces the company’s commitment to recognising and investing in its third-party distribution network.  Joseph Lim “The BFF program reflects our commitment to rewarding the partners who consistently deliver outstanding outcomes for Singapore SMEs. As a lender, we believe strong partnerships deserve meaningful recognition, and this program provides an opportunity to celebrate loyalty, performance and shared success through exclusive rewards and experiences,” he said. “SMEs are incredibly diverse, and their funding needs are rarely one-size-fits-all. We want Bizcap to be the first choice when businesses require fast, flexible capital, and partners play a critical role in making that happen. By rewarding high-performing partners, we’re encouraging greater reach into the SME community and helping more businesses access the funding solutions they need to thrive. “Our intermediary channel remains a key pillar of our growth strategy across Asia. We are committed to continuing to invest in our distribution network, and the BFF Program is another example of how we’re supporting partners who help drive value for their clients and for the wider SME ecosystem.” Gareth Tan, General Manager at Bizcap Singapore, said the program was designed to create a more rewarding partnership experience. Gareth Tan “At Bizcap, we’re always looking for ways to deliver a differentiated experience for our partners. The BFF program is our way of giving back to the brokers and referral partners who choose to work with us, while creating memorable experiences that reflect the value we place on those relationships. We want every interaction with Bizcap to feel rewarding, and this program takes that commitment one step further.” The launch of the BFF program marks the latest milestone in Bizcap’s growth journey in Singapore, building on the lender’s strong first year in market and continued investment in supporting both partners and SMEs with accessible funding solutions. Existing partners can contact their Bizcap relationship manager to learn more about the BFF program, while new brokers can find out more about partnering with Bizcap here.      Featured image: Edited by Fintech News Singapore, based on image by alionaursu via Magnific The post Bizcap Singapore Launches BFF Program to Recognise High-performing Partners appeared first on Fintech Singapore.

Bottomline has launched a CFO Suite that links treasury, payments and finance workflows as companies look for more control over cash flow and AI use. The modular platform brings together treasury cash forecasting, invoice processing, outbound payments, collections, dunning and cash application across the cash lifecycle. The suite is designed to help CFOs improve visibility over cash, working capital and risk while applying AI within controlled finance workflows. Bottomline CFO Suite is powered by the company’s BEA Agentic Platform, which it describes as a finance-first AI orchestration engine that works within existing finance workflows. The suite connects payment data, bank activity, invoices, ERP data and finance workflows into a single operational view. This is intended to help finance teams track cash as it moves through the business, improve forecasting, prioritise work and manage exceptions. Craig Saks Craig Saks, CEO of Bottomline, said, “Finance leaders are being asked to move faster, manage risk more tightly, and show where AI can create value, but they are often doing that with disconnected systems and delayed data. Bottomline’s CFO Suite is designed to close that gap by connecting the workflows that determine cash, working capital, and risk, while giving teams a controlled way to apply AI inside the processes they already rely on.” The suite supports cash forecasting, invoice processing, payments, collections and cash application. It also uses AI to support classification, data extraction, anomaly detection and payment matching. Colin Swain Colin Swain, Global Head of Product Solutions at Bottomline, said, “For CFOs, using AI in finance is only as useful as the data behind it, and only as trusted as the controls around it. Using the BEA Agentic Platform, Bottomline’s CFO Suite applies AI in a way that supports measurable outcomes, while keeping the auditability, explainability, and oversight finance teams rely on to trust the result.” The CFO Suite is available as a modular offering, allowing organisations to start with one priority area in treasury, accounts payable, accounts receivable or payments before expanding over time.     Featured image: Edited by Fintech News Singapore, based on image by mdsahazahan via Magnific   The post Bottomline Launches CFO Suite to Address Cash Flow Gaps appeared first on Fintech Singapore.

Adyen has agreed to acquire Orb in a US$335 million deal that will add enterprise billing capabilities to its payments platform. The transaction will be funded entirely from Adyen’s available cash resources and remains subject to customary closing conditions. Upon completion, Orb will become an indirect wholly owned subsidiary of Adyen. Orb’s co-founders will reinvest a meaningful portion of their proceeds into newly issued ordinary shares in Adyen. Founded in 2021 and based in San Francisco, Orb helps enterprises manage real-time usage data and complex pricing contracts. Its customers include Vercel, Glean, Replit and Supabase. Adyen plans to connect Orb’s billing infrastructure with its payments and risk systems. The company expects this to help merchants link pricing decisions with payment performance, fraud risk and transaction success rates. Adyen will manage Orb under an incubator model during the first phase of the acquisition. Orb will continue to support multi-payment service provider environments. Over time, Adyen plans to bring billing and payments into a more unified infrastructure for enterprise merchants. Ingo Uytdehaage Ingo Uytdehaage, Co-CEO of Adyen, said, “The structural complexity of modern billing has become the kind of infrastructure problem Adyen is built to take on. Helping customers optimise beyond the transaction itself has been an important part of our long-term direction, and recent moves have expanded our role further into the enterprise monetisation stack. Combining Orb’s billing product with Adyen’s payments platform closes the loop between what merchants charge and how those charges perform, enabling merchants to automate smarter revenue decisions in real time.” Alvaro Morales Alvaro Morales, CEO of Orb, said, “We built our architecture to process complex consumption logic at the event level, giving merchants total flexibility over their pricing frameworks. By joining forces with Adyen, we can connect this ingestion layer directly to real-time financial health signals, closing the loop between billing logic and payment success.” Adyen expects the Orb acquisition and its proposed acquisition of Talon.One to close on 1 July 2026, subject to regulatory approvals and customary closing conditions.     Featured image: Edited by Fintech News Singapore, based on image by muravev via Magnific The post Adyen to Acquire Orb for US$335 Million in Enterprise Billing Push appeared first on Fintech Singapore.

RippleX has launched the XRPL AI Starter Kit, a set of tools for developers building agentic payment applications on the XRP Ledger. The Ripple developer arm focuses on tools and infrastructure for the XRP Ledger ecosystem. The starter kit is aimed at AI agents that can make transactions, pay for services and settle value without direct human input. The first phase includes documentation, tools and integrations to help developers discover, learn and build agent-powered applications on XRPL. It also supports X402-powered payments using XRP and Ripple USD, or RLUSD, allowing AI agents to pay for API calls, AI model inference, compute and other digital services. XRPL Targets Agentic Payment Use Cases The launch comes as AI agents begin handling tasks that require payments and settlement, creating demand for infrastructure with fast settlement, predictable costs and automated transaction flows. The XRP Ledger settles transactions in around three to five seconds and supports predictable fees, native payments, multi-currency transactions and a built-in decentralised exchange. The starter kit includes an XRPL Docs MCP Server for MCP-compatible clients such as Claude Code, Claude Desktop, Cursor and custom agent frameworks. It also includes the XRPL Agent Wallet Skill and XRPL Payment Skill for Claude, covering wallet creation, balance checks, payments and transaction tracking. Developers can access new guides on xrpl.org, including a tutorial for completing an agentic transaction and a reference hub for building agentic payment flows. Future phases of the starter kit are expected to be shaped by developer feedback and emerging agentic payment use cases.     Featured image: Edited by Fintech News Singapore, based on images by nampix and zidan4ek via Magnific The post RippleX Targets AI Agent Payments With New XRPL Developer Kit appeared first on Fintech Singapore.

RippleX has launched the XRPL AI Starter Kit, a set of tools for developers building agentic payment applications on the XRP Ledger. The Ripple developer arm focuses on tools and infrastructure for the XRP Ledger ecosystem. The starter kit is aimed at AI agents that can make transactions, pay for services and settle value without direct human input. The first phase includes documentation, tools and integrations to help developers discover, learn and build agent-powered applications on XRPL. It also supports X402-powered payments using XRP and Ripple USD, or RLUSD, allowing AI agents to pay for API calls, AI model inference, compute and other digital services. XRPL Targets Agentic Payment Use Cases The launch comes as AI agents begin handling tasks that require payments and settlement, creating demand for infrastructure with fast settlement, predictable costs and automated transaction flows. The XRP Ledger settles transactions in around three to five seconds and supports predictable fees, native payments, multi-currency transactions and a built-in decentralised exchange. The starter kit includes an XRPL Docs MCP Server for MCP-compatible clients such as Claude Code, Claude Desktop, Cursor and custom agent frameworks. It also includes the XRPL Agent Wallet Skill and XRPL Payment Skill for Claude, covering wallet creation, balance checks, payments and transaction tracking. Developers can access new guides on xrpl.org, including a tutorial for completing an agentic transaction and a reference hub for building agentic payment flows. Future phases of the starter kit are expected to be shaped by developer feedback and emerging agentic payment use cases.     Featured image: Edited by Fintech News Singapore, based on images by nampix and zidan4ek via Magnific The post RippleX Targets AI Agent Payments With New XRPL Developer Kit appeared first on Fintech Singapore.

Shopee has started job cuts affecting hundreds of developers worldwide, representing about 8 percent of its developer team, as parent company Sea reviews staffing needs during a wider AI push. The reductions started earlier this week and include quality assurance positions, Bloomberg reported, citing people familiar with the matter. The report also indicated that Shopee may not be done reducing headcount. Sea, which operates Shopee, Garena and Monee, has been putting more resources behind AI as it searches for new growth engines. Chief executive Forrest Li has positioned AI as a key part of Sea’s long-term growth ambitions. Within Shopee, AI is already used to personalise shopping results and support merchant-facing tools. In February, Sea partnered with Google to develop AI tools across its businesses, including AI-powered shopping features for Shopee. Union Support Begins for Affected Workers A Sea spokesperson said the company adjusts headcount from time to time based on business needs and would provide assistance to employees affected by the changes. Singapore’s Creative Media and Publishing Union received advance notice of the planned job reductions. The union also deployed representatives to Shopee’s Geneo office on 8 June and Sea’s One North headquarters the following day to help affected employees. The Taskforce for Responsible Retrenchment and Employment Facilitation acknowledged Sea’s early engagement with the union and its cooperation with CMPU on support for affected staff. The affected employees are expected to leave the company in phases from late June to late August. The company has also committed to retrenchment payouts that follow Singapore’s Tripartite Advisory on Managing Excess Manpower and Responsible Retrenchment.     Featured image: Edited by Fintech News Singapore, based on image by Sea The post Shopee Job Cuts Affect Hundreds of Developers as Sea Expands AI Focus appeared first on Fintech Singapore.

Visa is introducing new capabilities and expanding its platforms to support automated, AI-driven digital commerce. Announced at the Visa Payments Forum 2026, the updates focus on building core frameworks for autonomous transactions and modernising backend payment infrastructure. To support agent-driven commerce, Visa introduced Visa Intelligent Commerce, an architecture designed to provide trust, controls, and connectivity for AI agents. As part of this ecosystem development, Visa announced a collaboration with OpenAI to enable secure payments using Visa infrastructure within agentic commerce environments. Jack Forestell “AI is transforming the front end of commerce. Stablecoins are reshaping the back end,” said Jack Forestell, Chief Product and Strategy Officer at Visa. “Visa’s role is to enable it to work securely, reliably and at global scale, for every participant in the ecosystem.” To help businesses adapt, Visa introduced Agent Score, a readiness evaluation tool for agentic commerce, alongside the Agentic Directory, a verified registry of legitimate merchants and agents. The network also deployed its Large Transaction Model, an AI system trained on billions of data points to improve fraud detection while reducing false declines. On the backend, Visa is expanding stablecoin settlement pilots across multiple blockchains, currencies, and regions. Building on earlier pilots, the network has moved billions of dollars in stablecoins across VisaNet, reaching an annualised run rate of approximately US$7 billion as of March 2026. This enables issuing banks to settle onchain seven days a week, with plans to extend this capability further across the ecosystem. The payment network is also building infrastructure to help banks convert traditional deposits into programmable, always-on tokenised deposits. Visa continues to expand its stablecoin-linked card programs, with more than 160 initiatives live or in development globally. For financial institutions, Visa is offering modular infrastructure through its cloud-native Pismo core banking platform. It also highlighted Unified Checkout, an orchestration layer designed to support both card and non-card payments within existing merchant systems.     Featured image credit: Edited by Fintech News Singapore, based on image by yelosole via Magnific The post Visa Unveils AI Commerce, Stablecoin Settlement and Token Infrastructure Upgrades appeared first on Fintech Singapore.

Mastercard has launched a new service that lets businesses use AI agents and software systems to make automated payments within set rules and limits. Agent Pay for Machines by Mastercard is designed for transactions that happen programmatically in the background of digital commerce. These include high-volume, low-value payments between AI agents, machines and software systems. The service is designed to allow these payments to be permissioned, orchestrated and settled at machine speed across Mastercard’s global payments network. In one example, an AI agent could set up an online store by buying a domain name, hosting service, images and checkout tools within a set budget. These transactions differ from traditional card or point-of-sale payments because they are automated, continuous and executed between systems rather than initiated one by one by a person. Jorn Lambert Jorn Lambert, Mastercard’s Chief Product Officer, said, “Agent Pay for Machines will create the conditions for a superbloom of AI business models. Machine payments can make it possible for services to be bought and sold among agents at fundamentally different scales than payments today — very high volumes, very small values, very fast and at extremely low latency.” Agent Pay for Machines builds on Mastercard’s Agent Pay programme, introduced in 2025 to support payments involving trusted AI agents. The new service supports credentialing, permissioning, transaction controls and settlement across multiple payment types, including cards, accounts and stablecoins. Businesses will be able to set spending limits and authorisation rules that are enforced programmatically. Verified participants will also be able to transact across providers and systems, with settlement supported across multiple rails. More than 30 companies are among the first participants and supporters of the service, including Adyen, Ant International, BVNK, Checkout.com, Cloudflare, Coinbase, Getnet by Santander, Global Payments, Lovable, OKX, Stripe and Tempo. Mastercard is working with partners to test use cases, develop common rules and support adoption across industries.     Featured image: Edited by Fintech News Singapore, based on image by thanyakij-12 via Magnific The post Mastercard Enables Businesses to Use AI Agents for Automated Payments appeared first on Fintech Singapore.

Thunes is expanding its US footprint with a new New York office to support cross-border payouts for American businesses. The new hub, located in Union Square, adds to Thunes’ existing US operations in San Francisco and Atlanta. The company is using capital from its US$150 million Series D funding round announced last year to support its expansion in the market. Thunes provides cross-border payment infrastructure for enterprises including gig economy platforms, marketplaces, payment service providers and financial institutions. Its network connects members to billions of endpoints across more than 140 countries and 90 currencies. Thunes said its US expansion is supported by its 50 money transmission licenses. Simon Nelson Simon Nelson, Chief Commercial Officer at Thunes, said, “The opening of our New York office is a pivotal moment for our team, but it is just one piece of a much larger picture. Our investment into the US market this year is a direct response to the massive demand from American businesses looking for faster, more transparent ways to move money globally.” Kyle Rosen Kyle Rosen, Head of Americas at Thunes, said “What truly sets Thunes apart in the US is that we own and operate our licenses in every state. This total control over our regulatory footprint is a huge differentiator for us and guarantees the highest level of reliability and speed. Our growth in New York is a testament to the hard work of our global team and our commitment to building a world-class presence in the U.S.” Thunes expects the New York office to support regional growth, closer industry engagement and hiring. The company is currently recruiting for several roles, with a focus on business development.     Featured image: Edited by Fintech News Singapore, based on image by DevStudio_D via Magnific The post Thunes Opens New York Office as It Expands US Payments Business appeared first on Fintech Singapore.

Grab is reportedly in discussions to join a funding round for Atome Financial, as the company looks to deepen its exposure to buy now, pay later and consumer lending in Southeast Asia. DealStreetAsia reported that the round could raise more than US$100 million, citing people familiar with the matter. The structure has not been finalised and may proceed in tranches. Atome Financial is part of Singapore-based Advance Intelligence Group and operates BNPL platform Atome and Indonesian digital lender Kredit Pintar. Lending Push Continues The talks come as Grab grows its lending and banking operations across consumers, merchants, drivers and SMEs. The segment remains loss-making. In the first quarter of 2026, Grab’s financial services unit recorded an adjusted EBITDA loss of US$17 million, while its gross loan portfolio reached US$1.44 billion, up from US$625 million a year earlier. Grab expects the unit to reach breakeven in the second half of 2026. The company has also expanded through deals. GXS Bank completed its acquisition of Validus Capital, the Singapore business of SME lender Validus Group, in April 2025. Grab also signed an agreement earlier this year to acquire US-based digital investing platform Stash Financial. Advance Intelligence Group last raised US$80 million in May 2023 from investors led by Warburg Pincus and Northstar Group, bringing its total funding to more than US$700 million.     Featured image: Edited by Fintech News Singapore, based on image by Grab The post Grab Reportedly in Talks to Join Atome Financial’s Funding Round appeared first on Fintech Singapore.

DBS will launch tokenised physical gold for customers in Singapore, with retail access expected through DBS digibank in the second half of 2026. The DBS Physical Gold Tokens will allow customers to access, hold and trade gold-backed tokens through a single platform. The bank described the service as a way to widen retail access to physical gold, which has traditionally been more readily available to institutional and accredited investors. Each token will be backed by one gram of physical gold held by DBS in a dedicated vault in Singapore. Customers will also have the option to redeem their tokens for physical gold. DBS is exploring plans to list the token on DBS Digital Exchange for accredited investors and institutional partners, with further details to be announced later. The offering is intended to make physical gold investment more accessible by allowing customers to buy smaller fractions of the asset in token form. Transactions are expected to be available around the clock, with near-instant settlement supported by blockchain technology. DBS will tokenise, issue, distribute and manage the gold tokens in-house. Li Zhen Li Zhen, Head of Foreign Exchange, Precious Metals and Digital Assets, Global Financial Markets at DBS, said, “Our ability to deliver DBS Physical Gold Token is built on DBS’ fully in-sourced, end-to-end institutional capability – physical vaulting, tokenisation engine, digital custody and digitised distribution. Through the secure and compliant deployment, underpinned by robust risk management and governance, we are able to minimise operational friction and streamline physical gold access and management for our clients.” The launch adds to DBS’ existing gold investment products in Singapore, which include funds and physical bullion. James Tan James Tan, Group Head of Investment Product and Advisory at DBS, said, “While our retail investors have been able to buy gold funds, access to physical gold has been largely available to only institutional and accredited investors. DBS has offered physical gold investments to wealth clients since 2013, and we are now leveraging tokenisation to broaden access, enabling more retail customers to invest in gold in a safe and meaningful way.”     Featured image: Edited by Fintech News Singapore, based on image by halalstock via Magnific The post DBS to Launch Tokenised Physical Gold for Retail Customers in Singapore appeared first on Fintech Singapore.

Conversations revolving around digital asset finance often return to the blockchain, but Shaun Chen’s concern sits closer to the point where value actually moves. Behind every transaction sits a private key, and Shaun, AVP, APJ Advisor – Quantum and AI Security at Thales, sees that key as the authority behind the asset. When someone compromises a private key, the institution faces more than a security breach because a validly signed transaction can move value before anyone fully understands what went wrong. Security around those keys helps determine whether digital asset infrastructure can move from experimentation into real commercial activity. Banks and fintechs may build new models around tokenisation or blockchain-based settlement, but those models still depend on how they protect the keys behind those services. Shaun makes a clear distinction between ordinary credentials and private keys. A password can be reset, while a compromised private key can authorise the movement of assets, making software-based handling harder to defend as digital asset activity scales. The risk also changes once institutions move beyond pilots. Controlled projects may run within a narrow environment, while production places more systems around the signing process and puts more pressure on internal controls to hold. Shaun’s answer points toward hardware-based protection. Institutions trying to secure digital asset keys need a clear separation between the systems that request a transaction and the infrastructure that protects the private key. The Limits of Keeping Private Keys in Software Software-based key storage may feel convenient during early experimentation because it helps teams move from concept to testing without adding too much operational weight. Shaun’s point, however, is that production changes the conversation. Once digital asset activity moves into real commercial use, the surrounding environment becomes harder to control. Private keys can become exposed when they are handled too closely by software systems, especially when access rights, workloads and operational processes begin to expand. Loose administrator controls, malware and more targeted cyber attacks can then turn software-based key handling into a serious vulnerability for institutions managing these assets. Shaun draws a firm boundary around that risk. Shaun Chen “Private keys should not be exposed to software environments.” His view reframes how institutions should think about securing digital asset keys. Strong application security remains necessary, alongside a key protection model that prevents private keys from being extracted in usable form when the surrounding software environment comes under pressure. A storage decision therefore becomes a control decision. If a private key can be taken out of its environment and used elsewhere, the institution has not really secured the authority behind the asset. Hardware Changes the Security Model Hardware security modules, or HSMs, address the software exposure problem by moving sensitive cryptographic work into dedicated hardware. Inside an HSM, institutions can generate and use cryptographic keys without handing them over to the wider software stack. The benefit becomes especially clear when institutions think about digital asset key security, because the private key can remain inside the HSM instead of sitting close to the application environment that supports the transaction workflow. Shaun described the distinction clearly when he said, “The private key does not need to leave the HSM to sign a transaction.” These systems still need to use the key because transactions must be signed before value can move. An HSM allows that to happen without giving the application direct possession of the private key. When the application sends a signing request, the HSM performs the cryptographic operation inside its secure boundary. After producing the signature, the HSM returns it to the transaction workflow while keeping the private key inside the device. In practice, a compromised application server should not automatically expose the private key, and system administration should not create a route to key extraction. Cryptographic authority stays inside dedicated hardware rather than spreading across software environments. Transaction Signing With the Key Kept Inside Blockchain transaction signing makes private key protection visible because the institution still needs to authorise movement without exposing the key that makes authorisation possible. A typical HSM-backed flow starts when the hardware module generates the keys. When a transaction needs a signature, the transaction data moves to the HSM, and the institution can route the request through policy and authorisation checks before the HSM signs it. After signing, the HSM returns the signed transaction to the application or transaction workflow, which can then submit it to the blockchain network. The private key stays inside the secure boundary throughout the process. The flow may sound simple, but it addresses one of the biggest risks in digital asset finance. Blockchain transactions often become difficult, or impossible, to reverse once someone validly signs and broadcasts them. An attacker who obtains the private key and signs a transaction may cause damage before anyone detects it. Shaun’s responses make clear that HSMs play a specific role in this chain. They protect the signing authority itself and keep the most sensitive part of the process away from ordinary software exposure. Control Over Key Usage Becomes the Real Test HSMs often get described as secure storage for keys, but Shaun’s answers point to a broader role. “The goal is not simply to store private keys securely,” he explained. Banks and custodians also need a clear record of how their systems use those keys. At the point of signing, the institution needs to know who triggered the request, whether the request followed policy and whether the sequence can be proven later. HSMs become part of the digital asset key security model when they place stronger controls around signing authority and transaction authorisation. A mature setup can enforce governed access, approval rules and key usage controls while keeping the private key away from the wider software environment. Such controls matter because they connect digital asset security directly to operational risk and governance. A bank needs to answer questions that go beyond cyber defence. Who initiated the transaction? Who approved it? Which key did the system use? Did the transaction follow policy? Did the system record the signing event? Can the institution prove what happened later? Shaun framed this around the need to govern the way an institution uses a private key, rather than treating protection as the end of the story. As these services involve more teams and counterparties, manual oversight becomes harder to rely on. This governance becomes even more critical as institutions begin exploring AI-driven financial workflows. When an autonomous agent or algorithmic system initiates a transaction at machine speed, manual oversight is impossible. The institution must rely entirely on the HSM to enforce policy, ensuring that every machine requested action is cryptographically authorized and bounded by strict hardware controls before value moves. Internal pilots may work with tighter human control, while regulated custody, tokenisation or settlement businesses need systems that can withstand audit, scale and scrutiny. Audit Trails Become Part of the Trust Layer Regulated institutions need a clear record of how they use private keys once these activities move into production. For institutions operating across fragmented regulatory environments in APJ, this level of control is also a matter of digital sovereignty. By anchoring private keys in dedicated hardware, an institution ensures it retains absolute, sovereign control over its assets and audit trails—regardless of which cloud environments or third-party platforms the surrounding applications run on. A signing event should show the key involved, the policy behind the approval, the authorised process that triggered it and the context around the decision. Without that record, an institution will struggle to explain how it controlled the authority to move value. Shaun’s answer places HSMs within the control plane behind institutional digital asset finance. HSMs help institutions determine how they generate keys, who can use them, which transactions they allow and what evidence they keep. The custody use case makes the issue easier to see. A custodian safeguards assets for clients, so these asset custody security ultimately depends on how well the institution protects and governs private keys. Once attackers gain control of the key, the custodian’s control over the asset is weakened. Tokenisation creates the same pressure around signing authority. As more financial instruments move into tokenised form, the systems behind those assets need stronger controls over actions that create, move or administer value. Those actions can carry direct financial consequences, so institutions need to govern them with the same seriousness as other high-impact financial operations. Hardware-rooted trust also gives institutions a model they already understand. Banks have long relied on HSMs in sensitive financial environments, and tokenised finance extends that model into a newer setting where value can move through programmable systems. Crypto Agility Moves Into the Planning Room The conversation does not stop with today’s private key risks. Digital asset systems still rely on cryptographic algorithms such as Elliptic Curve Digital Signature Algorithm (ECDSA) and Edwards-curve Digital Signature Algorithm (EdDSA).  Those algorithms remain widely used, but financial institutions are beginning to look further ahead as post-quantum cryptography enters long-term planning. Quantum computing may eventually threaten some public-key cryptographic algorithms used today. Moving every digital asset workflow to quantum-safe algorithms will take time because the wider blockchain ecosystem still has to mature around new cryptographic standards. Shaun’s answer avoids treating post-quantum readiness as an overnight switch. Institutions should avoid brittle key architectures that become difficult to govern or change later. The infrastructure behind signing authority should be treated as a long-term trust architecture rather than a one-off implementation. HSM-based architectures can support crypto agility by centralising key control and giving institutions a governed path to introduce new algorithms, hybrid mechanisms or updated signing approaches as standards mature.  C-level leaders should see the link between current key security decisions and future resilience. HSMs protect private keys now, while also helping institutions prepare for cryptographic change without rebuilding their trust architecture from scratch. Securing Digital Asset Keys Becomes a Strategic Decision Digital asset markets are maturing, and the infrastructure behind them has to keep pace. Blockchain protocols, custody frameworks and compliance processes will continue to matter, but weak private key protection can still undermine the whole model. Shaun brings the discussion back to the signing authority behind digital asset finance. Institutions that want to scale these services need infrastructure that protects that authority, governs its use and leaves evidence when questions arise. HSMs help provide that foundation by keeping the signing authority inside dedicated, tamper-resistant hardware. Cryptographic operations can still take place while the keys remain out of software environments, giving institutions a stronger base for governance today and crypto agility over time. As Shaun put it, “In digital asset finance, the private key is not just a credential. It is the authority to move value.” Institutions now have to protect that authority with infrastructure that can hold up as these tokenised asset finance becomes more institutional. Featured image: Edited by Fintech News Singapore based on an image by mrsiraphol via Magnific. The post Are HSMs Becoming the Standard for Institutional Digital Assets? appeared first on Fintech Singapore.

Generative AI is making fraud more convincing and easier to scale. Reports of Gen AI-enabled scams rose 456% between May 2024 and April 2025, while deepfake activity is also rising sharply in APAC. Banks and fintechs across APAC are under growing pressure to tighten fraud controls and strengthen identity checks. This webinar brings together fraud and risk leaders to discuss how financial institutions are responding, including the use of stronger biometric checks and advanced liveness detection. The growing impact of deepfakes in Singapore and across APAC Exploring advanced technologies as fraud prevention grows more complex Striking the right balance between biometric checks and a smooth customer journey How regulators in APAC are responding to AI-driven scam risks and rethinking legacy authentication Speakers: Catherine Paleracio, Chief Information Security Officer of Tonik Gan Kee Lim, Head, CyberSecurity & Tech Risk, GXBank Julius Rajeswaran, Chief Operating Officer, Ryt Bank Dominic Forrest, Chief Technology Officer, iProov Moderator: Izzat Najmi Abdullah, Senior Writer, Fintech News Network The post Webinar: The Deepfake Threat and What APAC Financial Institutions Are Doing About It appeared first on Fintech Singapore.

“Where is your key?” Asked in everyday life, the question might sound like a small inconvenience. When asked inside a financial institution handling digital assets, it becomes a test of whether the organisation truly knows how assets can be moved and protected. Ray Law, Senior Security Solution Architect at Thales, placed that question at the centre of any serious discussion on institutional digital assets, especially as banks and financial institutions move from experimentation into more operational use cases. As stablecoins, tokenised money and digital assets move closer to institutional workflows, private key control becomes central to how institutions move assets, protect them, audit activity and respond when something goes wrong. Much of the institutional conversation around blockchain and tokenised finance used to carry a cautious tone, with banks and regulators trying to understand where the technology could fit without weakening the trust that underpins the financial system. The conversation has since moved into more practical territory. Stablecoins are being considered for payments and treasury use, while banks are looking at tokenised forms of money and assets as part of a broader shift in how value could move through the financial system. As these use cases move closer to day-to-day financial operations, security becomes much more than a technical consideration. It becomes part of the business model. During the webinar, The Blueprint for Institutional Digital Asset Security at Scale, speakers from Visa, UOB Group, Rakkar Digital and Thales discussed how institutional adoption is changing, what risks are becoming more urgent, and why the next phase of digital asset growth will depend on more than technical controls alone. Stablecoins Are Moving Beyond the Crypto Conversation Sanchit Mall, Director of Digital Currencies, Asia Pacific at Visa, said the company has seen strong demand for stablecoin-related payments across both consumer and institutional use cases. “We have been seeing tremendous growth in the demand for stablecoin-related payments,” he mentioned. Consumers holding digital assets can link them to cards and spend through Visa’s acceptance network, while institutions are looking at stablecoins for money movement and treasury use. “Stablecoins can let you move money seven days a week instead of five days a week,” Sanchit said, adding that Visa’s stablecoin settlement activity had reached a run rate of almost US$7 billion this year. Stablecoin adoption is pushing the discussion beyond crypto as an investment product. Treasury use cases, especially those involving money movement outside traditional banking hours, make security harder to treat as a back-end technical issue. Institutions now have to show that they can control and monitor these use cases safely as digital assets become more embedded in financial operations. Sanchit framed it clearly. “It is no longer that the technology needs to be safe. It is also the governance. The technology and the governance should merge and tackle together.” Digital Assets Are Becoming Part of the Financial System Yip Kah Kit, Head of Blockchain and Digital Assets at UOB Group, made a similar point from the banking side, describing digital assets as having moved closer to the core of the financial system. UOB is looking at digital assets as part of a broader shift in how value moves, including tokenised money and investment products. Kah Kit was careful, however, not to frame this only as a technology build. Institutions also need to rethink the operating model around digital asset activity, especially how existing risk frameworks and internal processes cope with new forms of value movement. “We look at it from three perspectives. One is the infrastructure layer. Secondly, the processes and policies would need to be augmented. Last but not least would be the culture of the people,” he stated. His emphasis on culture carried through the wider discussion, especially as several speakers returned to the same underlying concern. Even with cryptographic controls and secure wallet infrastructure, the weakest point often remains human behaviour. The Weakest Link Is Still Often Human Arthit Sriumporn, Founder of Rakkar Digital, spoke from the perspective of a custodian who had held significant client assets. The company was founded in 2022, headquartered in Singapore, licensed in Hong Kong and Thailand, and at its peak had around US$700 million under custody. Holding that level of client assets made security much broader than technology, he explained. “It was not just technology in terms of security. In fact, most of the time technology is not compromised. People [are the ones who] compromise [it],” said Arthit. Rakkar responded by building multiple layers of control around its custody operations. Private key shards were kept on offline, air-gapped devices, while access to signing processes involved secure facilities and biometric checks. Client-side governance controls also determined transaction limits and approval requirements. Arthit also highlighted that a single transaction could require multiple people from both the client and custodian side before it could move forward. Layered controls also help institutions deal with attacks that begin away from the core system, such as a compromised employee device or a social engineering attempt. Rakkar’s founder shared one incident where one of the company’s developer devices was compromised around Token2049. The bad actor remained in the network for some time, but luckily, he said: “No asset got stolen. That is the real benefit of having multiple safeguards in place.” Private Keys Are Now Institutional Infrastructure Ray’s earlier question about where an institution keeps its key comes back into focus here. “The first question every institution needs to answer is, where is your key? How is it generated, stored and used?” Ray remarked. The question goes straight to the heart of digital asset security because private keys control the movement of assets. Once institutions start handling digital assets at scale, key protection becomes one of the most important parts of the institutional blueprint. Ray said insecure key storage and private key compromise remain among the root causes of major losses. Some institutions, he noted, still manage keys in software, databases or configuration files, which can create serious exposure. “Software keys can be copied and stolen silently,” said Ray. “By the time you know a key is compromised, the damage is already done.” Thales sees the answer in hardware-backed protection, policy-based access controls and a lifecycle view that covers how keys are generated, stored, used and recovered. Ray also stressed the need for multi-party authorisation, where no single person or compromised credential can move assets alone. “Beyond prevention, audit trail matters,” he added. A strong audit trail matters because institutions need to prove what happened, who approved an action, where it came from and whether controls operated as intended. Security Has to Move Beyond Perimeter Defence Kah Kit pushed the discussion beyond perimeter defence, arguing that institutions should not rely only on building higher walls around their systems as attack surfaces widen. Instead, they should consider an “assumed breach” approach. Institutions still need to secure their perimeter, but they also need visibility into activity inside their ecosystem. “In addition to securing your perimeter, we should also look at how we use transaction analytics or even AI to monitor what is happening within our ecosystem,” Kah Kit stated. Digital asset security also cannot just depend on one perfect wall. Institutions must be alert and need to always assume that something, somewhere, could fail, and design controls that can contain the damage before it reaches the assets they are trying to protect. AI Is Becoming Both the Threat and the Defence AI brings its own complications. Like blockchain, it can strengthen both attackers and defenders. Bad actors can use it to make fraud and social engineering move faster, while institutions can use AI-driven analytics to review activity at a scale that would be difficult for human teams to manage on their own. “The timeless truth is that we should always use technology to fight technology,” said Kah Kit. Yet he also warned against relying on automation alone. In his view, human oversight remains critical, especially when AI agents begin operating inside financial workflows. Kah Kit also raised the need for “cognitive brakes.” If AI agents are given authority to act within financial operations, institutions need clear boundaries and circuit-breaker-style checks before any mistake scales too quickly. Ray then brought the discussion back to the security of AI systems themselves. Institutions offering AI-powered tools must also protect those systems from manipulation, including attacks that try to trick models into exposing sensitive data. AI security, in other words, has to cover both sides of the equation: how institutions use AI to monitor threats, and how they protect the AI tools they deploy. Compliance Is Only the Starting Point Regulation is also evolving, but the panel made clear that compliance alone cannot be the ceiling. Ray said regulatory frameworks often mandate outcomes, such as securing private keys, without always specifying the exact technical model. “Do not just wait for regulations to tell you the answer,” he told. Institutions still need to treat regulation as a baseline and think carefully about whether their controls can withstand real scrutiny. After a breach, the harder question will be whether the institution did enough. Kah Kit, who previously worked with the Bank for International Settlements, offered a more regional perspective. He responded by saying that regulators in this part of the world have generally done a good job balancing trust, resilience, innovation and market development. But he also observed a shift in regulatory scrutiny. Previously, having the right policy and process may have been enough. Now, quoting Kah Kit: “It is not just enough that you have a nice policy that you file in the cabinet. [Regulators would want to know] How well do you implement them?” Digital Asset Risk Does Not Stop at Borders Effectiveness becomes harder to prove when digital assets cut across sectors and borders because criminal networks do not operate neatly within one jurisdiction, but financial institutions and regulators often still do. Kah Kit said the industry needs stronger collaboration models, including public-private mechanisms and cross-border intelligence sharing. He pointed to examples such as Singapore’s COSMIC platform and Malaysia’s national fraud portal in the traditional finance space, but said similar mechanisms need to develop in the tokenised world. “Criminals operate in networks, and as financial institutions, we cannot operate in silos,” he said. Sanchit also noted that regional consistency matters. In fragmented markets, bad actors can exploit weaker jurisdictions or regulatory gaps. “Inter-regional, inter-country collaboration is very important,” he said. Interoperability also becomes important once digital assets move beyond isolated use cases. Sanchit gave his thoughts that institutions will need ways to work across different networks and digital asset ecosystems if they want these models to scale. Security Should Be the Reason Digital Assets Can Scale Taken together, the discussion showed that institutional digital asset security cannot be reduced to a single tool, control or compliance checklist. Private key protection, hardware-backed security and multi-party approvals all matter, but they need to sit within a broader operating model that also accounts for governance, people and real-time monitoring. The institutions that move ahead will be the ones that design security into the operating model from the start, rather than adding it after the product is built. Arthit summed up the opportunity well. “Do not let security be the barrier,” he said. His point lands neatly at the end of the discussion. Security should not slow institutional adoption by sitting outside the business. Done properly, it becomes the reason digital assets can move into the financial system with trust and scale. Watch the full webinar, The Blueprint for Institutional Digital Asset Security at Scale, right down below. Featured image: Edited by Fintech News Singapore based on an image by user4894991 via Magnific. The post Institutional Digital Assets Are Growing Up, And Security Must Grow Up With Them appeared first on Fintech Singapore.

Reap has tapped Sumsub to support onboarding and compliance checks as it expands its stablecoin-native cards and payments infrastructure beyond APAC. The Singapore-based fintech will use Sumsub’s platform to automate verification for business customers and end cardholders. The partnership is aimed at helping Reap manage onboarding across markets with different regulatory, anti-money laundering and counter-financing of terrorism requirements. Darryl Wan Darryl Wan, Head of Legal, Risk and Compliance at Reap, said, “Onboarding is often the first real experience a customer has with Reap and it needs to be fast, simple, and compliant no matter where they are located. With the right infrastructure in place, we can scale into new markets and give businesses seamless access to stablecoin-powered cards, payments and financial tools. We are delighted to partner with Sumsub to enhance our verification and KYC processes.” Sumsub’s platform lets Reap configure verification flows by customer type, market and risk profile. It also supports reusable KYC, which reduces repeated verification steps for users who have already completed checks through Sumsub. According to Sumsub’s internal data, one in three applicants has previously been verified on its platform before starting a new onboarding process, although many are still asked to upload the same documents again. Sumsub’s platform can verify more than 14,000 types of identity documents from over 220 countries and territories. Its Liveness Detection technology is also used to verify individuals during onboarding. Penny Chai Penny Chai, Vice President, APAC at Sumsub, said, “Our platform is designed to address fragmented regulatory environments across jurisdictions while enabling seamless user experiences through reusable digital identity. By verifying users once and extending that trust across the ecosystem, we help fast-growing fintechs like Reap reduce unnecessary onboarding friction without compromising on compliance.” Reap is exploring further automation of onboarding processes as it enters new markets and rolls out more products.     Featured image: Edited by Fintech News Singapore, based on image by mangpor2004 via Magnific The post Reap Taps Sumsub to Support Compliance Checks Across New Markets appeared first on Fintech Singapore.

Credit Bureau Singapore and Experian Malaysia are working on a cross-border credit reporting service to help lenders assess borrowers with financial footprints in both markets. The proposed service will allow individuals to apply for consented credit reports across Singapore and Malaysia. It is intended to help lenders assess credit risk more accurately and improve access to financial products for consumers who live or work across both markets. The companies will explore a framework for the secure exchange of individual credit information. This will cover each bureau’s role in handling applications, consumer consent, data protection, governance and commercial arrangements. The service could benefit workers, entrepreneurs and professionals whose financial records span both countries. It may also help lenders avoid treating cross-border applicants as new-to-credit customers when they already have verified credit histories in the other market. For financial institutions, the data could provide a fuller view of an applicant’s financial obligations. This may support digital underwriting, reduce cross-border fraud risk and help lenders assess customers with financial ties to both markets. Credit Bureau Singapore and Experian Malaysia noted that consumer consent, regulatory compliance and data protection will be central to the initiative. Both parties will work within their respective legal and regulatory frameworks and engage relevant authorities as the project develops. William Lim William Lim, Executive Director of Credit Bureau Singapore, said, “As individuals and businesses operate more seamlessly across Singapore and Malaysia, credit information systems must evolve to reflect cross-border realities. This collaboration represents an important step towards enabling more seamless and responsible access to credit for consumers.” Dawn Lai Dawn Lai, Chief Executive Officer of Experian Information Services Malaysia, added, “Trusted data collaboration is key to strengthening digital financial ecosystems. By working together, we aim to enhance financial inclusion, improve risk transparency and support sustainable growth across both markets.”     Featured image: (From left) Dawn Lai, Chief Executive Officer of Experian Information Services Malaysia and William Lim, Executive Director of Credit Bureau Singapore The post Credit Bureau Singapore, Experian Malaysia Plan Two-Way Credit Reporting appeared first on Fintech Singapore.

Nuvei is in advanced discussions to acquire cross-border payments firm Payoneer for roughly US$2.7 billion, according to people familiar with the matter, as reported by Reuters. The potential transaction aims to merge Nuvei’s merchant payment capabilities with Payoneer’s extensive global payout network. The proposed US$2.7 billion purchase price incorporates Payoneer’s cash reserves, placing the enterprise value at approximately US$2.3 billion. Nuvei could finalise an agreement in the coming days, though talks remain ongoing and plans could still change. An acquisition would grant Nuvei deeper access to emerging markets and large online marketplace clients, including Amazon, Walmart and eBay. The move reflects a broader trend among payment companies seeking scale through acquisitions, particularly in the cross-border and business-to-business sectors amid slower growth in traditional payment processing. Nuvei has actively pursued expansion since private equity firm Advent International, alongside Novacap and CDPQ, took the Montreal-based company private in a US$6.3 billion buyout in 2024. Payoneer holds a market capitalisation of around US$1.7 billion at the time of reporting. The company generates much of its revenue from emerging market businesses selling into the United States and Europe. The firm has exposure to geopolitical and trade-related risks stemming from tariffs and trade tensions between the US and China. Customers in greater China accounted for 34% of Payoneer’s revenue in 2025, according to company filings. While the company increased its revenue by 8% to US$1.05 billion last year, its net income dropped 40% to US$73.2 million due to declining interest income and rising operational costs.     Featured image credit: Edited by Fintech News Singapore, based on image by John Caplan via LinkedIn The post Nuvei in Advanced Talks to Acquire Payoneer for US$2.7 Billion, Sources Say appeared first on Fintech Singapore.

“We’re not ready for this to be very clear,” warned Rajay Rai, Chief Information and Operations Officer at Trust Bank, during a Singapore AI CxO roundtable hosted by Fintech News Network and Alteryx. His warning landed because the room was not debating some distant version of banking. AI is already working its way through financial institutions globally, and the examples around the table showed how quickly it is becoming part of daily banking work in Singapore. Trust Bank has live chatbots and fraud use cases, while Tyme has brought manual KYC down from around 24 hours to 22 minutes with the help of AI. UOB is currently exploring agentic AI across areas such as AML and KYC, and DBS is measuring AI use cases through economic value. Still, the discussion did not feel triumphant. As the roundtable moved on, the concern that kept coming back among the speakers was whether Singapore banks can turn AI adoption into real readiness across the structures that support it. All of them agreed that adoption alone will not be enough if banks cannot make governance more practical and train people for the work AI is starting to change. Customer trust also remains a live question among them, especially as AI moves deeper into real banking workflows across Singapore. @fintechnewsnetwork After 35 years in banking, Rajay Rai, CIO Trust Bank says the industry has never faced a pace of change like AI. And it isn’t ready. He explains why. Here’s Rajay Rai’s take on AI in banking. fintech Digitalbanking AI banking @Trust Bank Singapore ♬ original sound – Fintech News Network – Fintech News Network Singapore Banks Are Already Past the Starting Line The roundtable showed a market where the technology has already moved into real workflows, with leaders from incumbent banks, digital banks, payments firms and technology providers all comparing lessons from live deployments and internal experiments. Much of the banking conversation last year still revolved around generative AI and productivity tools. This year, attention is moving toward agentic AI, where systems can take on more of the work across a process. As those systems become more active, banks have to decide how much autonomy they are comfortable giving them, and where human judgement must remain firmly in place. Luckily, Singapore’s regulatory environment is moving in the same direction. IMDA launched its Model AI Governance Framework for Agentic AI in January 2026, while MAS announced an AI risk management toolkit for the financial sector in March 2026. Both developments point to a wider industry concern over how financial institutions should manage responsibility and oversight as AI becomes more embedded in daily work. The same concern is showing up beyond Singapore as well, with Gartner warning that more than 40% of agentic AI projects could be cancelled by the end of 2027 due to weak business cases or risk controls. The warning matched the tone of the discussion as the leaders around the table were not dismissing AI, and many were already using it. Their concern was however, whether the institution around the technology can adapt quickly enough. AI Is Turning More Bankers Into Builders One of the stronger themes from the roundtable was the rise of citizen coding inside banks, where AI is changing who gets to participate in the build process. “The coding language is English,” Rajay said. He was pointing to a reality where more people inside the organisation can now easily translate business needs into working outputs without waiting for every change to pass through a traditional technology queue. But Rajat Malhotra, CTO of GXS Bank, looked at the same change from another angle. “Humans are evolving into full stack humans,” he mentioned. “It’s no longer, oh, I am an engineer. I am an operations person. I am a backend engineer.” Rachel Freeman, Chief Growth Officer at Tyme, saw a similar change inside teams that would not usually be placed at the centre of technology transformation. She considers that one of the biggest changes in 2026 was the “unleashing” of people across the organisation to build and improve processes themselves. Staff in areas such as operational risk and internal audit are now using AI to digitise manual work and rethink how those processes should run. Céline Le Cotonnec, Chief Data and Innovation Officer at Bank of Singapore, described a similar move toward business users becoming more hands-on with data and prototyping. She elucidates some examples happening in the private bank, where bankers and assistant relationship managers are already querying data from a single source of truth, while teams across the organisation build applications and prototypes before handing them to technology teams All of this gives Singapore banks a wider pool of people who can solve problems closer to where the work happens, although their readiness for AI becomes harder to manage when more employees can build. When more people can build, who decides what is ready to scale? More Output Means More Pressure on Senior Reviewers The speakers also believe that productivity gains in software development are already showing up inside banks, although the gains are also shifting pressure onto the people who have to review the work. Jackson Oh, CTO of ANEXT Bank, for instance, explained that the speed of engineering work had changed dramatically. “Our engineers are giving us feedback faster than the product guys can come up with,” he said. He told that junior engineers can now generate far more output with the help of AI, while senior reviewers on the other hand, need some time to review it properly. But Jackson noted that more code and faster iterations can only create value if experienced reviewers have the capacity to assess and approve the work. The review layer may therefore become one of the harder constraints as AI-enabled teams move faster. Banks can increase production speed, but senior judgement is much harder to scale at the same pace. Céline then pointed to the longer-term talent problem created by this new way of working. “There’s no job for fresh grad, but we need people with 20 years of experience to review codes and systems,” she uttered. Her comment captured one of the more uncomfortable workforce issues in the discussion. If early-career roles become thinner, banks may struggle to develop the people who later need to exercise judgement over AI-generated work. The industry now will have to rethink how it trains critical thinkers before the traditional apprenticeship path starts to weaken. AI Will Only Be as Good as the Bank Behind It A sharper point emerged as the discussion moved into software development and internal workflows. The faster banks want AI to work, the more discipline they need around what they feed into it. Rajay argued that productivity gains depend on the quality of the inputs before code generation even begins. “Your LLM is going to be a reflection of how good you are,” he noted. A bank with messy documentation or too many versions of the same process will see those weaknesses surface again in the model’s output. Other speakers arrived at the same issue through the data layer. Songhua Zhang from DBS Bank said one of the bank’s major initiatives this year is to tackle long-running data challenges, including the difficulty of giving data professionals reliable sources for modelling and reporting. Agentic workflows, in his view, could help because they can browse internal sources under proper access controls and make institutional knowledge easier to retrieve. Céline made a similar point from the wealth management side, where she believes that unstructured data and metadata quality matter if LLMs are expected to interpret information properly. Without that foundation, even a powerful model will struggle to produce output that teams can trust, because AI readiness for Singapore banks starts much earlier than deployment. A bank now has to make its own knowledge usable before it can expect a model to use it well. Governance Has to Move Into the Workflow Once AI becomes part of daily banking work, governance has to sit closer to the tools people actually use. GXS Bank representative, Rajat, described about building governance into its AI platform so employees do not have to make every judgement on their own about whether they are using the technology correctly or exposing information to the wrong tools. Revolut’s Raymond Ng also stressed the need for first, second and third lines of defence, especially when younger teams have powerful AI tools and more room to make decisions. Céline added another important point where she pointed out that control functions must become more technical. “We can’t have people doing AI governance that don’t know what they’re talking about,” she warned. Her point speaks to the practical side of AI readiness for Singapore banks. Governance cannot sit only in policy documents if the people reviewing AI use do not understand how the systems work. Banks still need formal policies, but those controls have to show up in the tools and review processes that teams use every day. Ross Morpeth, Director of Customer Experience for EMEA and APAC at Alteryx, brought the conversation back to trust and accuracy. Data quality still matters because, as he noted, even the smartest model becomes worthless if it is trained or fed with poor data. He also warned that the trust problem cuts both ways. “There’s also a danger that people trust it too much, and just take what AI tells them as gospel,” he said. The point lands because much of the AI debate in banking tends to focus on whether people will trust the technology enough. Overtrust may become one of the quieter risks for banks using AI because a confident answer can still be wrong, and in a regulated environment, speed is not much help if people stop questioning the output. Will Banks Still Look Like Banks? As the conversation moved from current use cases to the future of financial institutions, the discussion became less about AI tools and more about what kind of bank survives in a more programmable world. Rachel Freeman pointed out that Tyme is already thinking about who the financial institutions of the future will be, and how smaller players can avoid being displaced by newer, more programmable banking models. Vincent Fong, Chief Editor of Fintech News Network and moderator of the roundtable, then pushed the idea further and asked whether the future bank could be built with 10 or 15 people rather than hundreds. The idea sounds extreme at first, yet it follows naturally from the wider discussion. AI changes the economics of building and running financial services, which means the future bank may not look like a slightly more efficient version of today’s institution. Atul Bhuchar, Head of Transaction Banking Product APAC at SMBC, offered one of the most memorable ways to describe that choice. “Are we aspiring to be a faster caterpillar, or are we aspiring to transform into a butterfly?” he asked. Incumbent banks may find that question harder to avoid as AI moves deeper into the institution. Faster processes and better productivity can take banks only so far if AI begins to change the structure of banking itself. Banking Is Still a Trust Business Even after the discussion moved into 10-person banks and more programmable financial institutions, Arun Muraleedharan brought the conversation back to why banks may still matter. “Ultimately, banking is a trust business,” he said. Customers may welcome faster journeys and smarter tools but when money is involved, they still expect a bank to stand behind the decision, because AI can speed up the work without carrying institutional responsibility on its own. AI’s larger role will force banks to define trust more carefully. Banks will need clearer lines around human judgement and accountability when the technology influences decisions. After all, if banking is still a trust business, what does trust look like when more of the work is being done by machines? There was far more in the room than could fit into one article. Watch the full roundtable to hear how Singapore’s banking leaders are thinking through AI, trust and the future shape of financial institutions. The post Singapore Banks Are Moving Fast With AI, But Can They Actually Keep Up? appeared first on Fintech Singapore.