Skip to content
Submit incident
AII-1002-15053
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.Supporting sources are pending verification.

Taiwan banks' AI fraud detection freezes thousands of accounts

Taiwan banks deployed AI fraud-detection systems in April 2025 that froze thousands of legitimate customer accounts based on false positive pattern matching, affecting daily financial functions such as mortgage and tuition payments. The Financial Supervisory Commission responded by directing banks to improve AI precision and provide faster account unlock services by September 2025.

April 2025 Β· Taiwan domestic banks Β· Safety and wellbeing
April 2025
AII-1002-15053
Quick facts
Category
Safety and wellbeing
Country
Taiwan
Date
April 2025
Drafted by
Raidu Incident Detection

What happened

Taiwan domestic banks began deploying AI systems in April 2025 to detect fraudulent accounts and freeze suspicious transactions. The AI flagged accounts for unusual activity patterns such as sudden large deposits, frequent small transactions, or anomalies in transaction timing compared to established account behavior.

Thousands of legitimate customers reported their accounts were frozen without clear cause, including salary transfer accounts used for regular bills and loan payments. Customers faced freezes lasting approximately three months and reported being denied access to funds for essential expenses such as mortgage payments and school fees.

Reported impact

Public complaints escalated, with lawmakers reporting constituent grievances in the Financial Committee. Customers described account freezes as disruptive to daily financial life, and some reported being required to provide accounting records at bank counters to resolve freezes.

The Financial Supervisory Commission reported that in early 2025 warning account numbers (警瀺戢) surged to 150,000 from a prior level of 66,000, prompting the AI deployment as a control measure.

Index analysis

The AI system triggered false positives on normal account activity, such as fixed monthly salary deposits that varied slightly or occasional larger transactions. The system lacked sufficient granularity to distinguish between legitimate customer behavior and fraud indicators.

Governance controls

In response to public outcry, the Financial Supervisory Commission in September 2025 directed banks to improve AI model precision, establish dedicated unlock telephone lines, and provide 24-hour account unlock services. The commission also required banks to authorize phone-line staff to resolve simple cases within one hour.

Classification

Category
Safety and wellbeing
Industry
Financial and insurance services
Country
Taiwan
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Economic or property
Harmed
Consumers, General public
Failure modes
Misidentification
AI systems
Bank AI fraud detection system

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.

How sources are verifiedSuggest a source for this incident