UK Financial Conduct Authority (FCA) joined forces with research and technology partners Turing Institute and Plenitude Consulting to create a synthetic dataset designed to improve the fight against money laundering. This initiative addresses a long-standing obstacle to financial crime prevention: the difficulty of accessing realistic transaction data without compromising customer privacy or violating strict rules. data protection rules.
money laundering It remains a massive global problem, with estimates suggesting criminals move 2 to 5 percent of the world’s GDP (approximately $800 billion to $2 trillion each year) through legitimate financial channels.
Traditional rules-based detection systems often suffer because suspicious activity often involves multiple accounts, entities, and transaction models rather than occurring as isolated incidents.
Banks and regulators have clashed in the past legalEthical and technical hurdles when trying to share or analyze real customer data to develop better tools.
Even anonymised records may lose critical behavioral signals or be at risk of re-identification.
To overcome these obstacles FCA The Alan Turing Institute collaborated with Plenitude Consulting and Napier AI on the Synthetic Data and Anti-Money Laundering project.
The team started with anonymised real-world transaction information from UK retail banking. Using advanced generation techniques, including Adaptive and Iterative Mechanism (AIM), they produced a completely synthetic collection of customer profiles, accounts, and customer profiles. transactions.
This data set deliberately incorporates a number of realistic money laundering typologies, such as structuring, while closely reflecting the statistical characteristics of real banking activity. payments rapid layers of funds between linked accounts, cyclical “round trip” movements, and high-risk cross-border transfers just below reporting thresholds.
Differential privacy Controls are placed throughout to ensure that no individual or specific process can be reverse engineered.
Initial evaluations confirmed that the dataset maintains high statistical fidelity to the original source material, preserves the complex relational patterns required for detection tests, and successfully accommodates detectable laundering scenarios of varying complexity.
While it cannot replicate every unknown criminal tactic, it provides a secure, shareable environment that closely replicates real market conditions.
Project partners contributed complementary strengths: Turing Institute brought deep expertise in privacy-preserving synthetic data methods, Plenitude offered expert financial crime knowledge and Napier AI provided practical technology experience in detection systems.
FCA provided regulatory oversight and strategic direction. This synthetic source will now be released through FCA’s Digital Sandbox platform.
It forms the center of upcoming Synthetic Data AML Solution Sprint, where participating companies can test and demonstrate innovative sensing technologies, especially those supported by artificial intelligence, without ever handling live customer information.
Applications for the Sprint will close on April 26, 2026. The exercise is more effective, data-focused compliance tools level the playing field for smaller innovators and generate valuable evidence for future regulatory approaches.
The project takes a significant step forward in the UK’s wider data collection strategy by demonstrating that synthetic data can serve as a reliable, privacy-first alternative to real datasets. technology against economic crimes
Editors We predict this will foster faster trials, stronger collaboration across the industry, and ultimately a more resilient financial system that better protects customers and markets from illicit finance.
The dataset is intended to be a complementary resource rather than a complete replacement for operational resources. dataThere are regular update plans to keep up with evolving threats.
