Chinese money laundering networks (CMLNs) are no mystery to the financial crime community, yet the money keeps moving.
According to research from Consilient, which recently delved into CMLNs, FinCEN’s recent Financial Trend Analysis identified more than 137,000 Bank Secrecy Act filings between 2020 and 2024, representing over $300bn in suspicious activity linked to these networks.
The uncomfortable question, Consilient argues, is not whether this activity is being reported, but how much of it is only recognised after funds have already passed through the system.
CMLNs do not rely on simple, isolated transactions. Their activity is multi-step and non-linear, layered through trade-based schemes, bulk goods purchases, real estate and cash-intensive businesses, supported by money mules and informal value transfer systems. Individually, the transactions are structured to look ordinary, spread across accounts, institutions and time. The risk only becomes visible when activity is viewed in sequence.
That is precisely where traditional AML controls fall short. Rules-based monitoring assesses transactions against predefined thresholds and generates alerts on individual events, not on how behaviour connects across a network. Investigators end up with fragmented visibility, reviewing alerts in isolation without a clear picture of how related activity develops. By the time a pattern is recognised, the funds have often already moved.
Consilient highlights three structural weaknesses. First, static rules cannot keep pace with adaptive networks, which continually adjust transaction flows, account usage and timing; a rule built for one typology will miss the next variation. Second, detection is fragmented across institutions, with each bank seeing only a slice of activity that appears low risk in isolation. Third, red flags do not capture behaviour: risk emerges from how transactions relate to one another across accounts, entities and time, not from any single event.
Effective detection, the firm suggests, requires a shift towards behavioural patterns, recognising how funds cluster, sequence and link across steps. This also sharpens precision. Many transfers to China are entirely legitimate, and a behavioural lens helps investigators separate routine cross-border flows from genuine laundering risk, cutting false positives.
This is where federated machine learning changes the model. Each institution trains a model on its own data locally, and only the learned parameters are shared and combined. No transaction-level data leaves the institution, yet the resulting model reflects patterns observed across multiple datasets. The approach builds a collective view of risk across the industry while respecting privacy constraints, and the models can be retrained as CMLN methodologies evolve, keeping detection aligned with how these networks actually operate.
For more, read the full story here.
Copyright © 2026 FinTech Global
Copyright © 2018 RegTech Analyst
