The State Office for Combating Financial Crime (LBF NRW) and the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, together with the Lamarr Institute for Machine Learning and Artificial Intelligence, are developing a prototype to leverage AI in tax investigation. The project aims to enhance the analysis of digital evidence, enabling faster detection of suspicious transactions and financial flows linked to terrorism financing.
The scientific foundations for this innovative use of AI in tax investigation are based in part on research carried out at the Lamarr Institute. Lamarr focuses on high-performance, trustworthy AI methods designed to address real societal challenges – including security-critical applications such as financial crime detection.
“As part of this project, we are creating an AI system that can, for instance, recognize photos of invoices and convert them into text. This makes it possible for investigators to ‘chat’ with the evidence and search specifically for relevant information,” explains Prof. Dr. Christian Bauckhage, Co-Director of the Lamarr Institute and Lead Scientist for Machine Learning at Fraunhofer IAIS. “AI helps tax investigators work through large volumes of data efficiently, identifying key indicators of transactions related to terrorism financing.”
Supported by funding from the state government’s security, migration, and prevention program, the project demonstrates how modern technologies like AI in tax investigation can play a critical role in protecting society. Finance Minister Dr. Marcus Optendrenk emphasizes: “Our goal is to uncover and disrupt financial flows that may support terrorism. The money trail can lead us to the source of this threat – and we will use every modern tool of investigation to safeguard our society.”
The six-month collaboration involves developing, implementing, and documenting the AI prototype. The system integrates retrieval-augmented generation technology, allowing investigators to filter and analyze data in various formats and efficiently access crucial information.