FCA Synthetic Data and Anti-Money Laundering project report: Key points for financial services firms

The FCA has developed a fully synthetic dataset containing embedded money laundering typologies, which will be made available to firms participating in the upcoming Synthetic Data AML Solution Sprint through the FCA Digital Sandbox. The dataset is designed to allow firms to test and demonstrate new AML detection approaches – including AI-driven tools, without needing access to real customer data.

‍

‍

The report is clear that synthetic data should complement, not replace, live operational data. Firms that treat synthetic data as a definitive representation of real-world risk run the danger of over-reliance, neglecting the ongoing real-world calibration and validation that effective AML systems require.

‍

The dataset incorporates well-established money laundering typologies such as structuring, layering, circular transaction patterns, and high-risk jurisdiction transfers, alongside more complex variations. However, the report acknowledges that the dataset cannot capture the full spectrum of financial crime, as it can only reflect typologies that have already been identified and codified - meaning emerging or novel laundering techniques will not be represented.

‍

The payment may be delayed for no longer than is necessary to make checks of the payer or third parties to determine whether the payment should be executed. The maximum delay to executing the payment is D+4 from receipt of the payment order.

The FCA warns that synthetic datasets risk becoming outdated if they do not evolve alongside the threat landscape, as financial criminals continuously adapt their techniques in response to technology, regulatory pressure, and enforcement actions. The report notes that without regular refreshes incorporating emerging typologies and intelligence feeds, firms may end up optimising their systems for yesterday's risks rather than tomorrow's.

‍