Fraunhofer Institute for Industrial Mathematics ITWM
Fraunhofer Institute for Industrial Mathematics ITWM
Geld / 100 Euro-Schein löst sich in digitale Pixel auf
© iStockphoto
Projekt »I-WASH«: Künstliche Intelligenz gegen Geldwäsche –Verdachtsmeldungen automatisiert analysieren

AI Against Money Laundering

In the »I-WASH« project, we are developing AI methods for the automated analysis of suspicious money laundering reports. This enables law enforcement agencies to detect patterns more quickly, prioritize cases more efficiently, and reduce the workload in investigative processes.

© iStockphoto

Artificial Intelligence Against Money Laundering – Automatically Analyzing Suspicious Activity Reports

Project »I-Wash«: Money Laundering in Focus – Ai-Supported Algorithms, Graph-Based Methods, and Chatbot Assistance for More Efficient Investigations

The number of suspicious transaction reports (STRs) in Germany has been rising sharply for years. Legislative changes since 18 March 2021 and the EU-wide Anti-Money Laundering package are further tightening the requirements for law enforcement agencies. As a result, the police forces of the federal states face the challenge of manually analyzing highly heterogeneous and extensive datasets – requiring significant personnel resources and leading to lengthy investigative processes. In the collaborative project »I-WASH«, we are working together with the Berlin Police and IFS, a subsidiary of T-Systems, to develop practical AI methods for the automated analysis of STRs, faster pattern recognition, and targeted support for investigators.

The project is funded under the Federal Ministry of Research, Technology and Space (BMFTR) funding guideline »Applications in Civil Security« and is intended to close existing capability gaps in the fight against money laundering.

From Case-by-Case Review to Pattern Recognition

The focus is on methods that efficiently analyze both individual suspicious reports and larger interconnections.

First, AI-supported models reduce the workload in case-by-case processing: they automatically extract key information from reports, assess it, and assist in generating structured reports. This results in standardized decision-making foundations that enable better case prioritization.

In addition, we are developing methods that analyze STRs at scale and make cross-case patterns visible. Graph-based machine learning approaches link individuals, accounts, and transactions into networks in which suspicious structures – such as complex layered payment flows or recurring participants – become clearly identifiable.

Ai-Supported Assistance Systems for Investigations

Another component is a specialized chatbot based on a large language model (LLM). It answers case-related questions, explains domain-specific background information, and supports navigation within extensive data sets. This provides investigators with an interactive assistant that combines expert knowledge with data access.

With »I-WASH«, the aim is to create a combination of modern AI methods, graph-based pattern recognition, and interactive assistance that fundamentally improves the handling of the growing number of suspicious money laundering reports. Feedback from users at the police and the IFS is continuously incorporated into the development process. This ensures that the resulting algorithms are not only technically powerful but also fit seamlessly into the operational workflows of investigative work.

Transferable AI Solutions for Other Security and Financial Actors

The algorithms developed in the project form the basis for further industrial research and development projects – ranging from the advancement of anomaly detection systems and integration into investigative and specialist software to the optimization of specialized chatbots.

The solution approaches are deliberately designed to be transferable to other stakeholders and domains. In addition to law enforcement and police authorities, banks, insurance companies, and healthcare institutions are particularly relevant fields of application. The methods for individual case analysis and cross-report pattern detection can be adapted for nationwide situational awareness and other areas of crime – for example, fraud in the healthcare sector. Furthermore, they open up potential for fraud detection in other institutional contexts, such as network analysis in social and economic data or anomaly detection in large databases.

Project Duration, Funding and Partners

The project »I-WASH« is funded by the German Federal Ministry for Research, Technology and Space (BMFTR) under the funding initiative »Applications in Civil Security«. The aim of the project is to develop AI-based methods for the automated analysis of suspicious activity reports (SARs) related to money laundering, thereby supporting law enforcement agencies in handling growing data volumes and increasingly complex investigations.

We work closely together with our project partners, the Berlin Police and IFS, a subsidiary of T-Systems. The Berlin Police contributes its operational expertise in money laundering investigations, while IFS provides its competence in the development of security-critical IT systems. Our team at Fraunhofer ITWM is primarily responsible for developing AI-supported algorithms for the automated analysis of SARs, graph-based methods for pattern detection, and intelligent assistance systems based on large language models (LLMs).

The developed solutions are designed to be transferable to additional security and financial application domains in the future. The project is scheduled to run until the end of 2027.

Related Links: