Criminal activities in the financial sector is undoubtedly one of the major challenges to the protection of the legal economy. In a social and economic context characterised by increasing and pervasive use of new technologies, criminal organisations are also increasingly using artificial intelligence to assist with money laundering.
The United Nations Office on Drugs and Crime (UNODC) estimates that money laundering — the set of activities aimed at concealing the illegal origin of money or other goods — represents between 2% and 5% of the global GDP, with a value ranging from 715 to 1,900 billion euros. These money flows allow criminal organisations to access substantial funds and invest in increasingly sophisticated technologies.
,In this context, financial institutions serve as a critical gateway allowing ‘dirty money’ to enter the financial system, so they are first in line in the fight against financial crime. Financial institutions are legally required to monitor their customers’ banking activities and report any suspicious transactions to regulatory authorities. However, traditional information systems to support transaction monitoring have their limitations. Since they are primarily based on predefined rules, they are unable to detect illicit transactions that fall outside established criteria.
Artificial intelligence is an indispensable tool in the fight against financial crime for several reasons. First, the quantity and complexity of global financial transactions have increased exponentially, making it impossible for humans to monitor and analyse them manually in an effective manner. AI can process huge volumes of data in a very short amount of time in order to identify suspicious patterns and anomalous behaviors that might indicate fraudulent activity or money laundering. Moreover, AI can learn and adapt continuously, thereby improving its ability to detect new types of threats as they emerge. This is crucial in a context where financial criminals are constantly developing new and increasingly sophisticated techniques to evade detection.
This is illustrated below with two examples of artificial intelligence applications used by a leading financial intermediary, one is for the monitoring of “negative news” and the other concerns money laundering schemes linked to “velocity”.
Negative news monitoring is an example of applying artificial intelligence solutions (in particular natural language processing) to provide automated analysis of unstructured data taken from news articles published in newspapers and other media. As part of their obligations to fight financial crime, financial institutions are required to monitor, among other things, news relating to their customers published in the press that might point to illegal conduct or money laundering. Obviously, analysing all press articles in various languages on a daily basis looking for negative news about customers is a complex and costly operation.
Thanks to artificial intelligence, it has become possible to automate this kind of analysis. The AI system processes huge amounts of news and unstructured data in various languages on a daily basis. It groups the information according to specific criteria (e.g. using a taxonomy of offences relevant to money laundering). The AI system extracts key information from the news article, such as the people and/or companies involved and their roles (e.g., suspect, arrested, etc.) in the context of the criminal events described. It then rates the correspondence between the customer and the news item using bespoke algorithms, taking into account any matches between the subjects mentioned in the ‘negative news’ and the bank’s customers, the relevance of the offence, and the role of the customer.
The phenomenon of “velocity”, on the other hand, is related to the need to identify complex patterns of transactions aimed at hiding the illicit origin of funds. One technique used by criminals is to divide large sums of money into many small, apparently legitimate transactions executed in a short period of time, thereby evading traditional checks based on static thresholds, a practice known as smurfing. How can smurfing be detected in a vast network of millions of financial transactions? One approach is to represent transactions on a graph network. It is then possible to efficiently identify subgraphs, i.e. segments of the network, that exhibit smurf-like behaviour. The evidence confirms that money launderers exploit different geographies and financial institutions to make it difficult to “follow the money”[1].
In conclusion, artificial intelligence is an essential tool in the fight against increasingly digital and sophisticated financial crime. Looking ahead, artificial intelligence will increasingly serve as a tool for detecting and responding to new criminal techniques and models of behaviour in real time. It will analyse large amounts of data from different sources with ever-increasing precision, with the aim of identifying anomalies and suspicious patterns.
In this context, generative AI is also taking on an increasingly important role. While still in a prototypical state, it is very promising, especially for extracting values from the multiple systems and documents that financial institutions already have in their possession. It could automate the most repetitive tasks, and it could help analysts be more efficient and effective in their activities.
[1] To learn more about relevant techniques and scientific results, please refer to the article: “Smurf-based Anti-Money Laundering in Time-Evolving Transaction Networks” by Michele Starnini, Charalampos E. Tsourakakis, Maryam Zamanipour, Andre Panisson, Walter Allasia, Marco Fornasiero, Laura Li Puma, Valeria Ricci, Silvia Ronchiadin, Angela Ugrinoska, Marco Varetto and Dario Moncalvo
