Taiwan’s financial sector is strengthening its defenses against fraud, with the Financial Supervisory Commission (FSC) stating that as of the end of May, 15 financial institutions have implemented facial recognition warning functions on approximately 1,600 ATMs. Observations from banking industry insiders suggest a 60% decrease in fraudulent withdrawal transactions at high-risk locations following the introduction of this feature. By the end of the year, further assessments will be conducted to consider expanding the implementation to more ATMs.
The Criminal Investigation Bureau of the Ministry of the Interior in Taiwan has noted that covering one’s face with a mask or helmet while making ATM withdrawals increases the difficulty of evidence collection and tracking. As a solution, the Bureau recommends the implementation of a mechanism that requires ATM users to show their faces. In December of last year, the FSC initiated discussions with banks on the feasibility of introducing facial recognition warning functions on ATMs. The initial stage involves using warning sounds without the ATM refusing withdrawals from customers.
According to the FSC’s explanation to the Central News Agency, by the end of May of this year, 15 financial institutions had implemented facial recognition warning functions on around 1,600 ATMs. The FSC has requested the banking association to gather feedback from the pilot institutions to formulate recommendations for further actions, with plans to report to the FSC by the end of August.
Cathay United Bank has emphasized the use of AI technology to enhance ATM fraud prevention capabilities. Their focus includes prioritizing high-risk locations for deployment, real-time detection of abnormal behaviors, and fraud prevention alerts, constructing a robust defense line against fraudulent activities. In addition to detecting facial coverings, other fraud prevention measures have been implemented, such as detecting frequent card changes, mobile transaction alerts, and anti-peeping reminders, with continuous improvements in AI recognition and alert design.
Since the introduction of AI-based ATM fraud prevention mechanisms by Cathay United Bank in 2025, approximately 1,000 AI-equipped ATMs had been deployed by the end of May. Monitoring of actual operations will determine adjustments and potential expansion based on changes in risk profiles and operational outcomes. Observations have shown a 60% decrease in fraudulent withdrawal transactions and no recurrence of fraudulent withdrawals at locations previously prone to such activities following the trial of various technological fraud prevention measures.
Regarding the initial implementation of AI-based ATM fraud prevention mechanisms, Cathay United Bank acknowledges that some customers may feel uncomfortable temporarily removing face coverings such as masks, helmets, or sunglasses during transactions. The bank continues to adjust and optimize processes. Observations from actual transactions indicate that the impact of these reminders on normal transactions is minimal, with the majority of customers successfully completing transactions even after reminders.
Cathay United Bank’s subsidiary, Cathay United Bank, has introduced AI facial recognition technology into their ATMs, automatically detecting if users have their faces covered. The feature has been deployed in high-risk fraud prone locations reported by the police, with functionality going live in mid-April. When customers use the ATM and do not fully expose their faces, they will be prompted by both visuals and voice commands to remove any face coverings, allowing the transaction to proceed after fully revealing their faces or waiting for 120 seconds.
Statistics from Cathay United Bank have shown that around 80% of customers conduct ATM transactions without any face coverings, while about 20% have some form of covering. Among customers with face coverings, 95% voluntarily remove them upon seeing the prompt, while the remaining 5% proceed after waiting 120 seconds. The bank plans to continue evaluating policies and aims to increase the number of ATMs equipped with this feature.
E.Sun Bank has incorporated fraud prevention features using AI facial recognition and smart detection technologies on selected ATMs. By doing so, they effectively reduce the risk of criminal groups using covering items to avoid identity verification and withdraw illicit funds. As of July 15th, E.Sun Bank has installed facial recognition warning functions on around 150 ATMs, accounting for about 3.5% of their total ATMs. The bank aims to reach 400 ATMs with this feature by the end of this year, increasing its coverage to 9.4%.
Yongfeng Bank, by implementing an “AI Facial Recognition” system in the automated service area of its ATMs, allows customers to enter and access ATM services by showing their full face, preventing dishonest individuals from obscuring their features to make withdrawals. As of the end of June, facial recognition access control has been installed in 72 branches of Yongfeng Bank, with plans to increase installations continuously before the year-end, surpassing a 60% adoption rate. Additionally, AI facial recognition cameras will be set up on non-bank automated devices, with over 40 expected to be completed by the end of the year.
In response to the implementation of facial recognition warning functions on ATMs, the Ministry of Finance recently stated that public sector banks have achieved a 2% implementation rate as of the end of April, estimating a 30% adoption rate by the end of this year.
(Translated from Central News Agency)