AI-Powered Romance Scams and Voice Phishing Defraud Users in Taiwan and Japan
AI-powered scams targeting romance and financial interactions reportedly spread in Taiwan and Japan by combining AI-generated emotional messaging with phishing and voice impersonation to defraud victims. In Japan, a hybrid voice phishing scheme exploited 155 bank customers in two days, while Taiwan reports romance scammers using AI to manage dozens of emotional relationships simultaneously to extract money.
What happened
Romance scams using AI-generated messaging systems reportedly operate by impersonating marriage brokers, successful professionals, and overseas executives on social media and dating platforms. The AI systems generate natural-sounding messages, remember personal details, and provide emotional comfort to build trust before requesting money for investments, medical emergencies, or travel expenses.
In Japan, a voice phishing scheme combining automated voice calls, real customer service impersonators, and AI-generated phishing emails began spreading in November 2025. Scammers pose as banks via automated voice systems, transfer calls to human actors, obtain email addresses under the pretense of online banking updates, then send AI-generated phishing emails with malicious links to fake banking websites to steal credentials.
Reported impact
Taiwan: Romance scammers using AI can simultaneously manage dozens of romantic relationships with minimal staff by deploying AI to monitor multiple conversations, exploiting loneliness and emotional vulnerability for financial gain.
Japan: Hokkaido Bank alone reported 155 customer complaints within two days in late 2025. Police note the hybrid approach combining voice interaction with AI-generated emails defeats traditional detection methods based on unnatural language, as AI now generates naturalistic emails in Japanese that evade human scrutiny.
Index analysis
The scams exploit AI's ability to simulate emotion, remember user details, and adapt communication style to individual targets. Past romance frauds failed due to stiff language and cultural errors; current AI systems mimic local language, understand slang, and adjust tone based on responses to create idealized romantic partners.
Voice phishing succeeds partly because human voice interaction combined with professionally written emails lowers victim defenses compared to text-only scams. Police analyze that victims lower their guard during friendly voice guidance then comply with link-clicking instructions, defeating the wariness typically applied to unexpected suspicious emails.
Governance controls
Taiwan: Experts call for platforms to label AI interactions clearly so users know they are communicating with machines. Recommended measures include building anomaly detection for templated emotional responses, pre-transfer risk warnings, and enhancing digital literacy and emotional manipulation detection in education.
Japan: Financial institutions stress they never request account credentials by phone or email, and recommend customers avoid clicking unfamiliar links and report suspicious activity to police or financial institutions immediately.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.