Taiwan’s Federated Banking AI Platform and Regulatory Framework for Countering Deepfake Financial Fraud

Photo Caption:
Anya Schiffrin, Co-Director of the Technology Policy & Innovation Concentration at the School of International and Public Affairs, Columbia University, visited Taiwan AI Labs in January 2026. Following her visit, she reported on Taiwan’s Federated Banking AI Platform and collaborative anti-fraud initiatives in her article: Deepfake Financial Fraud (Data & Society).(https://datasociety.net/library/deepfake-financial-fraud/)

Abstract
The rapid proliferation of generative artificial intelligence has enabled increasingly sophisticated forms of financial fraud, including deepfake-enabled impersonation, investment scams, and cross-border money laundering. Taiwan has emerged as an important case study in developing a coordinated regulatory and technical response to this evolving threat landscape. This paper examines Taiwan’s cross-sector anti-fraud governance framework—linking digital platforms, telecommunications infrastructure, and financial institutions—and analyzes the role of federated artificial intelligence systems in strengthening detection and prevention. It further explores how collaborative initiatives such as the Taiwan AI Labs Infodemic platform and the Eagle-Eye Anti-Fraud Alliance contribute to a “cheapest cost avoider” model of distributed responsibility across the scam ecosystem. Taiwan’s experience offers a replicable policy and technical model for jurisdictions confronting the intersection of generative AI, deepfakes, and financial fraud. [10][11][9]

I. Introduction
The emergence of generative artificial intelligence and synthetic media has significantly transformed the landscape of financial fraud. Deepfake-enabled scams increasingly rely on coordinated use of digital advertising, telecommunications infrastructure, and global payment systems to deceive victims and rapidly move illicit funds across borders. Traditional regulatory approaches, often sector-specific and reactive, have struggled to keep pace with the scale and speed of AI-driven fraud.

Taiwan provides a notable case study in developing an integrated regulatory and technical response. Through comprehensive legislation, sectoral coordination, and federated AI-based detection systems across its banking sector, Taiwan has adopted a whole-of-system approach to combating deepfake financial fraud. [1][10][13]

II. Taiwan’s Regulatory and Policy Framework
2.1 Cross-Sector Anti-Fraud Architecture
Taiwan’s Fraud Crime Hazard Prevention Act (2024) establishes a unified legal framework explicitly designed to prevent misuse of financial systems, telecommunications networks, and digital platforms for fraud [1][2][10]. Rather than treating scams as isolated criminal incidents, the law recognizes fraud as an ecosystem phenomenon enabled by multiple interconnected infrastructures.

The framework integrates three key layers:

Online platforms and digital advertising systems
Telecommunications infrastructure
Payment and financial systems

This integrated model aligns with lifecycle-based fraud prevention:
Prevention and disruption: enforceable obligations across platforms, telecom providers, and financial institutions to reduce exposure and interrupt scams early.

Detection: institutionalization of risk-based monitoring, including high-risk telecom identification, financial alerts, and cross-agency intelligence sharing.
Accountability: strengthened penalties and sentencing thresholds for organized and large-scale fraud.
By regulating all major intermediaries simultaneously, Taiwan addresses the structural conditions that allow scams to scale.

2.2 Platform and Digital Advertising Governance
Taiwan’s Ministry of Digital Affairs (MODA) has implemented regulations targeting scam advertising as a primary entry point for financial fraud [3][4][5][12][15]. These measures treat digital advertising infrastructure as a high-risk vector requiring proactive oversight.

Key regulatory components include:
mandatory verification of advertiser identity and funding sources;
platform anti-fraud compliance plans and periodic transparency reporting;
a legally defined 24-hour takedown requirement for identified scam advertisements; [15]
standardized verification and reporting frameworks effective from November 2024 [4][15].

This model operationalizes a “platform duty of care,” linking advertising funding sources to distribution infrastructure and reducing the monetization window available to fraudulent campaigns. It represents a concrete application of platform accountability frameworks discussed in contemporary digital governance literature.

2.3 Telecommunications Governance
Telecommunications networks are treated as a critical enabling layer in fraud operations. Taiwan has introduced risk-tiered controls that include:
restrictions on re-registration for users whose telecom services were terminated for fraud-related activity;
strengthened identity verification and monitoring of telecom accounts;
enforceable penalties for telecom providers failing to comply with anti-fraud obligations [1][10].

This risk-tiered telecom governance complements platform and financial oversight, embedding preventive controls within communications infrastructure rather than relying solely on individual vigilance.

2.4 Emergency Network Intervention Mechanisms
The Act authorizes emergency measures allowing authorities to request internet service providers to restrict access to scam infrastructure, such as suspending DNS resolution or limiting access to fraudulent domains in urgent cases [1][10].

While such mechanisms can rapidly reduce harm, they require procedural safeguards, transparency, and oversight to balance effectiveness with civil-liberty protections. Taiwan’s model illustrates how emergency intervention tools can be integrated into broader regulatory frameworks.

2.5 Financial Sector Coordination and Data Sharing
Taiwan’s financial regulators have advanced cross-sector anti-fraud coordination mechanisms that position banks as critical gatekeepers. These include:
real-time identification and protection of suspected scam victims;
cross-industry verification between banks and virtual asset service providers (VASPs);
development of cross-institution analytics platforms for detecting suspicious accounts and transactions;
joint interdiction and notification systems to interrupt fraudulent fund flows [6][7][13][14].

Such measures address key structural challenges in deepfake-enabled fraud, including rapid cross-border transfers, fragmented mule-account networks, and the speed of digital payment systems. They enable intelligence sharing and collective detection at scale.

2.6 Technology-Neutral Criminal Law Adaptation
Taiwan has also explored strengthening aggravated fraud provisions by recognizing the use of synthetic or manipulated media—such as AI-generated images and voice—as a method of committing fraud [8].

This reflects a technology-neutral legislative strategy, allowing existing criminal law frameworks to address emerging AI-enabled crimes without requiring entirely new statutory regimes.

III. Technical Contributions: Federated AI and Multi-Sector Collaboration
3.1 Infodemic.cc: Platform-Side Detection
The Infodemic platform developed by Taiwan AI Labs provides analytical infrastructure for detecting coordinated disinformation and scam campaigns across digital platforms. [16][22][23]

Key functions include:
early identification of scam advertisements and deepfake impersonation campaigns through risk mapping;
cross-platform narrative tracking to identify migration from public platforms to private messaging channels;
generation of measurable transparency metrics such as detection rates, response times, and repeat-offender patterns.

These capabilities strengthen regulatory enforcement by providing auditable evidence chains and operational metrics for platform accountability.

3.2 Eagle-Eye Anti-Fraud Alliance: Federated Banking Detection
Taiwan’s financial sector has begun deploying federated AI collaboration models across banks to detect and disrupt fraud at scale. The Eagle-Eye Anti-Fraud Alliance exemplifies a privacy-preserving collaborative detection framework. [11][13][9]

Key features include:
Privacy-preserving federated analytics
Financial institutions can detect coordinated fraud patterns without centralizing sensitive customer data, reducing privacy risks while enabling system-wide detection.
Mule-network detection
Cross-bank pattern analysis enables identification of fragmented transfer networks commonly used in scam operations.
Bank–VASP coordination
Collaborative risk-tagging, transaction monitoring, and step-up verification mechanisms support rapid intervention in high-risk transactions.

These mechanisms align with broader regulatory efforts to strengthen cross-sector intelligence sharing and financial safeguards [6][13][14].

IV. Discussion: The Cheapest-Cost-Avoider Framework
Taiwan’s approach reflects the “cheapest cost avoider” principle in regulatory design. Rather than placing primary responsibility on individual victims, the framework assigns measurable obligations to intermediaries—platforms, telecom providers, and financial institutions—best positioned to detect and prevent fraud at lower systemic cost.

This distributed responsibility model recognizes that effective mitigation of AI-enabled financial fraud requires coordinated intervention across the entire scam lifecycle, from initial exposure to final fund transfer.

V. Conclusion
Taiwan demonstrates that effective governance of deepfake-driven financial fraud requires integrated regulatory frameworks and collaborative technical infrastructure. By combining cross-sector legislation with federated AI-based detection systems across its banking sector, Taiwan has developed a scalable model for addressing the evolving risks posed by generative AI and synthetic media.

As deepfake-enabled financial fraud continues to expand globally, Taiwan’s experience offers valuable lessons for policymakers and financial regulators seeking to implement coordinated, ecosystem-level responses to emerging digital threats.

References
[1] Fraud Crime Hazard Prevention Act (Taiwan, full text)
https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=D0080226
[2] Fraud Crime Hazard Prevention Act – relevant provisions
https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=D0080226
[3] Ministry of Digital Affairs. Anti-fraud requirements for online advertising platforms and transparency reporting
https://moda.gov.tw/press/press-releases/14562
[4] Ministry of Digital Affairs. Sub-regulations on advertiser identity verification and transparency reporting
https://moda.gov.tw/press/bulletin/14534
[5] MODA. Regulatory details for online advertising platforms and transparency reporting
https://law.moda.gov.tw/LawContent.aspx?id=GL000158
[6] Financial Supervisory Commission. Cross-industry anti-fraud coordination and analytics platform initiative
https://www.fsc.gov.tw/ch/home.jsp?aplistdn=ou%3Dnews%2Cou%3Dmultisite%2Cou%3Dchinese%2Cou%3Dap_root%2Co%3Dfsc%2Cc%3Dtw&dataserno=202511130001&dtable=News&id=96&mcustomize=news_view.jsp&parentpath=0%2C2
[7] Executive Yuan. Anti-Fraud Action Guideline 2.0
https://english.ey.gov.tw/News3/9E5540D592A5FECD/faccc48c-1d4c-45c8-aa1b-73d9c283a73d
[8] Ministry of Justice. Amendment explanation on aggravated fraud and synthetic media
https://www.lawbank.com.tw/news/NewsContent.aspx?NID=192621.00
[9] Fubon Financial Holdings. Taiwan banking sector AI anti-fraud collaboration overview
https://www.fubon.com/financialholdings/news/news_1240816_786711.htm
[10] Ministry of the Interior (GLRS). Fraud Crime Hazard Prevention Act (English text)
https://glrs.moi.gov.tw/EngLawContent.aspx?id=795&lan=E
[11] Fubon Financial Holdings (English). The Expansion of the “Eagle Eye Anti-Fraud Alliance” (32 banks, Oct 16, 2023)
https://www.fubon.com/financialholdings/en/news/news_1231123_732397.htm
[12] MODA (Legal annex download). Identity verification technologies / transparency report format (annex file)
https://law.moda.gov.tw/Download.ashx?FileID=11123&id=82&type=ANNE
[13] Financial Supervisory Commission (English). Tech-driven prevention of fraud (FinTech PoC / alliance mention, Nov 13, 2024)
https://www.fsc.gov.tw/en/home.jsp?dataserno=202411130001&dtable=News&id=54&mcustomize=multimessage_view.jsp&parentpath=0
[14] Cathay Financial Holdings. Cross-institution federated learning anti-fraud PoC (Chinese)
https://www.cathayholdings.com/holdings/lastest_news/news_archive/newsarticle?newsID=BOnUVR84m0aX-ZpBAJvpGg
[15] MODA (Legal content). Notification period / “24 hours” requirement for online advertising platforms (effective Nov 30, 2024)
https://law.moda.gov.tw/EngLawContent.aspx?id=81&lan=E
[16] Infodemic.cc (Chinese homepage describing real-time monitoring and coordination detection)
https://infodemic.cc/zh-hant
[22] Infodemic.cc (English reports page)
https://infodemic.cc/en/reports
[23] Taiwan AI Labs / Infodemic.cc press release (PDF)
https://infodemic.cc/files/press_release_v1.pdf