Latvia’s AI Tool Cracks Down on Suspicious Cross-Border Payments
Latvijas Banka has introduced an innovative tool to bolster anti-money laundering efforts. The unsupervised machine learning model detects anomalies in cross-border payment data, aiding in the fight against financial crime. Since 2018, Latvia has seen a significant decline in suspicious payment flows with Russia, reflecting regulatory reforms and geopolitical shifts.
The tool, developed since 2018, employs the Isolation Forest algorithm to pinpoint isolated observations in payment data. It considers nine dimensions, including payment volume, economic ties, risk factors, and intrinsic payment characteristics, to determine the degree of anomaly. The model uses a calibration factor called the contamination index to estimate the prevalence of anomalies within the dataset.
The tool does not guarantee that anomalies indicate high-risk payments. Expert interpretation is crucial for accurate assessment. However, it significantly improves efficiency in analyzing potential money laundering threats. Since its implementation, anomalies in payments with Russia have declined, reflecting the impact of AML regulatory reforms and geopolitical shifts.
Latvia has made notable progress in mitigating and managing money laundering risk. The tool's ability to track changes in the overall level of anomalies over time provides valuable insights into evolving risks across the sector. Continuous supervision and expert interpretation are essential to maximize its effectiveness.
