Tainan, Taiwan’s sixth-largest city, is preparing to launch a pilot program in May 2025 aimed at reducing ATM fraud through the use of AI-driven facial recognition technology. The initiative, announced by Mayor Huang Wei-che on April 15, targets so-called “money mules”—individuals hired by fraud gangs to withdraw illicit funds while concealing their identities behind masks or motorcycle helmets.
Unlike earlier biometric ATM deployments in Taiwan—such as those by E.SUN Commercial Bank and Taishin International Bank—where facial recognition was used for customer authentication with pre-registered biometrics, the Tainan system is designed as a deterrent, not a gatekeeper. Users will receive voice prompts instructing them to remove face coverings before making a transaction. If the system cannot register a face within 10 to 15 seconds, an alarm will sound to alert nearby bank staff or bystanders. Crucially, the system will not block the transaction, focusing instead on visibility and psychological deterrence.
Meng Chih-cheng, head of Tainan’s Research, Development and Evaluation Commission, explained that current laws do not require individuals to reveal their faces at ATMs, a regulatory loophole that has made it easier for criminals to exploit anonymous withdrawals.
This initiative builds on Taiwan’s broader embrace of biometric technology in banking. In recent years, Taiwanese banks have increasingly adopted facial recognition for customer verification, often combining it with one-time passwords or PINs for enhanced protection. However, Tainan’s approach is distinctive in that it does not verify users against a database, but rather attempts to discourage fraudulent activity through passive surveillance mechanisms.
The pilot will initially be limited to a select number of ATMs in the city, and local authorities plan to closely monitor its impact on reducing ATM-related fraud. Depending on the results, the system could be expanded citywide or incorporated into national banking infrastructure, potentially prompting future regulatory updates.
Despite its innovation, the pilot also raises important questions around data privacy and public transparency. Officials have not disclosed how biometric data will be processed, stored, or protected—an increasingly critical issue as facial recognition systems become more prevalent. Civil liberties organizations in Taiwan have not yet issued formal responses, but similar deployments globally have triggered debate over surveillance, consent, and proportionality.
There are also concerns about practical limitations. For instance, legitimate users who wear masks or helmets for health, religious, or personal reasons may feel inconvenienced or unfairly singled out. Additionally, since transactions can still proceed even if a face is not recognized, the effectiveness of the system as a deterrent remains to be proven. Fraudsters may adapt by targeting ATMs that do not yet have the system installed or by employing more sophisticated disguises.
More broadly, Tainan’s pilot reflects a growing regional trend. In countries like Singapore and Vietnam, facial recognition is increasingly being used not only to deter fraud but to authenticate users—often replacing physical cards or PINs for high-value transactions. In those cases, access is denied if biometric verification fails, making them more stringent than Tainan’s deterrence-focused model.
If successful, the Tainan program could become a stepping stone toward stricter national standards or biometric requirements at ATMs. Until then, the pilot will serve as a test of whether visibility and psychological pressure can curb fraudulent withdrawals without infringing on user convenience or triggering public backlash.
Sources: Tech in Asia, Taiwan News, Focus Taiwan
—
April 16, 2025 – by Ali Nassar-Smith
