AI Advances Boost Fraud and Financial Crime Detection

Traditional methods of detecting financial crime are becoming increasingly insufficient in a data-saturated world. Financial institutions (FIs) worldwide face evolving challenges in combating financial crime, which grows in complexity and scale each day. This dynamic threat landscape underscores the need for advancements and innovations in detecting, preventing, and combating financial crime.

New AI Models for Financial Crime Detection

Wolfgang Berner, co-founder and CPO of Hawk, discussed the potential of large transaction models (LTMs) — generative AI models adapted to financial crime — in establishing robust detection and prevention mechanisms. Berner explained to PYMNTS that LTMs treat transactions as sentences, training the model on the language and grammar of transactions, similar to how GPT-4 is trained on web text.

“The model develops a very good understanding of transactions, how transactions relate to each other, and what is genuine or possibly suspicious,” Berner said. This training allows the model to excel at finding novel criminal activities and reducing false positives by detecting patterns and anomalies.

Advantages of LTMs Over Traditional Models

Berner highlighted two key advantages of LTMs: higher resolution and natural language understanding. LTMs maintain a broad view of transactions, understanding language data within transactions, such as reference texts indicating intent. This capability enables LTMs to detect complex relationships and patterns, providing an edge over traditional models.

Technical Foundation and Effectiveness

The attention mechanism, a key feature in both large language models and LTMs, helps the model focus on relevant transactions and elements within them. This enables LTMs to detect correlations between transactions, such as money mule behavior, while ignoring genuine transactions. LTMs work on an individual transaction level, maintaining high resolution and clarity, which traditional models often miss.

You can read how Hawk updated his AI platform to enhance financial crime detection.

Implementation and Results

Testing the efficacy of LTMs is crucial. Berner shared Hawk’s approach, emphasizing practical application and benchmarking against traditional models. One test compared Hawk’s LTM-based false positive reduction model with a traditional one, with the LTM one achieving a 30% improvement in false positive reduction while maintaining the same level of precision.

Future of LTMs in Financial Crime Prevention

Looking ahead, LTMs are poised to evolve further, addressing new challenges and use cases in financial crime prevention. Berner predicted that LTMs would serve as a foundational model, evolving through specific use cases and continually improving existing applications, where the LTM can act as a core infrastructure layer for new applications to be built atop. One development is using generative AI in screening for sanctions and identifying sanction circumvention through semantic matching.

“By understanding context and intent, the LLM can relate different attributes, enhancing our ability to detect sophisticated financial crimes,” Berner noted.

As financial criminals' strategies and tactics evolve, so must the technologies designed to thwart them, ensuring a more secure financial landscape.