Collaboration in anti-money laundering (AML) has long been a shared ambition across the financial services sector. Credit Reference Agencies already exchange payment data, fraud hubs in certain regions share intelligence, and industry forums regularly call for joint action. Regulators, too, reinforce the point: money laundering and fraud are collective threats that demand a collective response.
According to Consilient, some initiatives, such as the UK’s National Crime Agency’s (NCA) Data Fusion programme and the Netherlands’ TMNL, have sought to unite efforts. Yet despite widespread agreement, most projects stall at the pilot phase or falter when faced with operational, legal, and competitive realities.
Silos persist, promising utilities fizzle, and privacy concerns regularly halt progress. Institutions continue to duplicate investigations and onboarding checks, missing risks that a joint approach could uncover.
The desire to collaborate exists—banks, regulators, and compliance teams are generally supportive. But the frameworks, incentives, and regulatory clarity needed to make collaboration safe and practical remain absent. Until these change, collaboration will remain technically feasible yet practically unattainable.
Six main barriers explain why AML collaboration rarely scales. Regulatory misalignment means that, although the Financial Action Task Force (FATF) provides a global framework, local interpretations vary widely, making data-sharing initiatives legally complex. Data privacy laws such as GDPR and CCPA impose tight restrictions, while internal fragmentation sees risk, compliance, legal, and technology teams working in silos. Outdated or incompatible infrastructure also limits operational feasibility, and competitive pressures—combined with reputational concerns—discourage transparency. Finally, unclear governance and uncertain returns cause many joint platforms to lose momentum before launch.
Some projects have succeeded. Singapore’s AML/CFT Industry Partnership (ACIP) thrived due to regulatory support, a targeted scope, and privacy-conscious information sharing. The UK’s NCA Data Fusion programme has also made progress, embedding law enforcement in its design and focusing on high-impact typologies. In contrast, SWIFT’s Transaction Monitoring Utility failed to attract participation, Nordic banks’ shared KYC utility broke down over governance disputes, and TMNL in the Netherlands wound down after struggling with new EU AMLR requirements.
A new approach—federated learning—is offering a breakthrough. This model allows banks to train AML detection models locally, sharing only encrypted model updates rather than raw customer data. It preserves institutional control, broadens behavioural pattern detection, meets regulatory scrutiny, and scales easily as typologies evolve. Early adopters report significant gains, including 75% analyst efficiency improvements and quadrupled detection rates without increasing false positives.
The lesson is clear: collaboration must be private by design, outcome-focused, and operationally viable within existing compliance frameworks. Federated learning is emerging as one of the few models capable of delivering on AML’s long-promised shared response.
Copyright © 2025 RegTech Analyst
Copyright © 2018 RegTech Analyst
