Asia Pacific (APAC) is bearing the brunt of global fraud, according to a study by LexisNexis. The report highlights that the scam industry in the region continues to grow, reportedly employing hundreds of thousands of people, some under forced labour conditions. Human-initiated attacks surged by 61% year-on-year, while robotic (automated) attacks rose by 6%.
Japan experienced a significant rise in romance scams, investment fraud, and phone-related social engineering schemes. Hong Kong recorded the highest per capita fraud losses, while in Singapore, around 86% of police-reported scams involved authorised payment scams.
As fraud escalates across APAC, the need for more sophisticated detection strategies has become urgent, and AI is stepping in to meet the challenge.
“In APAC, with the rise in social engineering and authorised push payment (APP) fraud, traditional systems often fall short,” said Leo Li Shiwei, CEO of TrustDecision. “That’s where AI has begun to evolve.”
He explained that APP fraud is particularly difficult to catch because it often appears legitimate: the user logs in, verifies with OTP, and initiates the transaction themselves. “Traditional systems see a normal transaction. What they miss is the why behind it,” he said.
According to Shiwei, AI models are now trained to go beyond anomaly detection—they aim to detect intent. “They flag unusual behaviour: sudden changes in transaction size, skipping usual steps, or transfers made at odd hours to unfamiliar recipients. More advanced models go further, checking if the user is on a call while making the transfer—a red flag in many scam cases.”
Some models even interpret behavioral cues under stress, like rapid tapping, hesitations, or erratic scrolling. Others are linked to telco data to spot known scam numbers or repeated contact patterns.
Fraud isn’t always about stolen credentials anymore. It’s about manipulation. Leo Li Shiwei
“Fraud isn’t always about stolen credentials anymore. It’s about manipulation,” Shiwei said. “AI needs to understand context, behaviour, and pressure,especially in APAC, where mobile-first users and real-time payments create the perfect conditions for social engineering to thrive.”
Building trust with explainable AI
The rise of explainable AI (XAI) in fraud detection is another trend taking root in the region, driven by growing demand for trust, fairness, and compliance.
“APAC regulators are raising the bar on AI transparency, especially in markets like Singapore, Malaysia, and Australia,” Shiwei noted. “Financial institutions can’t just flag a transaction—they need to explain why. That’s pushing innovation in how fraud models are built and communicated.”
He described a shift from opaque, black-box models to hybrid systems that combine machine learning with rule-based decision trees, making outcomes easier to trace and audit. Some banks now use layered explanations: technical for risk teams, simplified for customer service, and regulatory-aligned for audits.
“XAI is also becoming real-time,” he added. “Instead of post-analysis reports, explanations are now delivered alongside the decision. This supports faster resolution and improves the customer experience.”
Shiwei emphasised that explainability isn’t just about meeting compliance; it’s about building trust. “In a region with rising digital adoption but uneven financial literacy, being able to clearly explain why a transaction was flagged can be the difference between keeping or losing a user.”
The APAC challenge: fragmented yet fast-moving
“APAC isn’t one market; it’s dozens, each with its own payment habits,” Shiwei said. “We’re evolving very fast, but not uniformly.”
QR payments in Southeast Asia are accelerating, with efforts to build joint national infrastructures. E-wallet use and cross-border transactions are increasing. However, the pace of digital adoption varies significantly across cities and countries.
That variation poses a challenge for traditional rule-based systems. “What looks suspicious in one market might be perfectly normal in another,” Shiwei said. “Consumers now have multiple payment methods at their fingertips and it all needs to be analysed cohesively to avoid false positives.”
AI, he explained, helps read these patterns at scale. “It learns local behaviours: how users top up, when they transact, what devices they use; and spots risks that rigid rules miss. That’s how banks and fintechs in APAC stay ahead, not by copying global models, but by training AI on how people really pay, borrow, and move money across this region.”
Still, deploying AI fraud detection at scale is not without hurdles—especially given APAC’s diverse regulatory environments.
“Regulations don’t move in sync,” Shiwei noted. In Indonesia, for instance, data must be stored locally. “If a bank uses a cloud-based AI model trained in Singapore, it hits a wall. It needs separate infrastructure, separate models, and local data pipelines just to stay compliant. Multiply that across 10+ markets, and scaling becomes complex fast.”
Shiwei also pointed out that technological gaps are another barrier. “Some countries have real-time payment rails and unified IDs. Others don’t. An AI system built for Singapore’s instant payments won’t work the same in the Philippines, where infrastructure is still catching up.”
Legacy systems complicate things further. “Many banks still process fraud checks in batches. Integrating real-time AI into that stack isn’t plug-and-play,” he said. “Scaling AI in APAC means building systems that flex with regulation, run in local environments, and integrate into uneven infrastructure. It’s not just about good models. It’s about making them deployable in the real world.”
Collaboration as a competitive advantage
In a region shaped by collectivist values, collaboration is proving key to overcoming the hurdles in AI-powered fraud detection. APAC banks and fintechs are working closely with telcos, e-commerce platforms, and government agencies to build shared, AI-driven ecosystems.
“No single player has the full picture,” Shiwei explained. “Banks see transaction data. Telcos see device and SIM behaviour. E-commerce platforms observe shopping patterns. Governments have identity and law enforcement data. To fight fraud that moves across channels, these players are starting to connect the dots.”
He advocates for a shared intelligence network with clear rules of engagement. “Banks and fintechs should integrate telco signals like SIM swap history and call activity into their risk models. Telcos can flag suspicious usage patterns. Platforms can provide behavioral risk signals. Regulators must set standards for privacy, data exchange, and coordinated response.”
Real innovation isn’t just in the tech. It’s in the trust and governance that make this collaboration work. Leo Li Shiwei
“What ties it all together is AI,” he concluded. “It’s the only way to process these fragmented signals fast enough to detect real threats. But the real innovation isn’t just in the tech. It’s in the trust and governance that make this collaboration work.”
“Fraudsters already operate as networks. Now the ecosystem is starting to fight back the same way,” he concluded.
