The threat of financial crime is becoming more and more prevalent, and Financial Institutions (FIs) are challenged by the increasing sophistication of crimes committed by money launderers and fraudsters. The United Nations Office on Drugs and Crime (UNODC) estimates that between 2-5% of global GDP each year stems from laundered or illegally obtained money 1. Increasing digitalisation poses opportunities both for criminals, but equally for FIs, allowing them to create robust lines of defence. In this article, we will examine the threats posed by clients operating in high-risk industries and how these risks can be mitigated, particularly with the use of digital and Artificial Intelligence tools.
Compliance teams need to take extra care when dealing with clients that operate in so-called ‘Cash-Intensive Businesses’ (CIBs). Cash transactions are often favoured by criminals because record-keeping can more easily be dispensed with, because the electronic proof of these transactions doesn’t exist. A few recurrent examples of CIBs are: Casinos & Gambling, Construction, Charities and NPOs (Non-Profit Organisations), Money Service Operators (e.g. currency exchanges) and Cryptocurrency industries.
FIs still rely on processing numerous physical documents for both onboarding and Know Your Client (KYC) reviews. As of 2025, the main method used is ‘periodical KYC reviews’, where clients are reviewed at specific time intervals, and which require a high degree of resources (e.g. manual KYC completion, collection of physical documents). This article discusses why a transition towards ‘Perpetual KYC’ 2 is a pertinent step for FIs to take.
Certain FIs still face a lack of advanced data analytics tools, meaning that they might not spot complex deviations in the transactional behaviour of clients. AI-enhanced Transaction Monitoring tools are highly effective at flagging potential threats caused by clients who work in high-risk industries. Companies like Fico 3 and SAS 4 offer AI-powered tools for transaction monitoring which can analyse vast amounts of data, detect patterns and/or anomalies that humans could not detect by themselves, as well as be used on an ongoing basis. There are other digital suites which offer a more holistic Financial Crime protection, one of the best examples being FinCense developed by the company Tookitaki 5 .
Client identity verification is a crucial component in all due diligence procedures and FIs still often rely on manual methods here as well, e.g. human verification and certification of ID cards. AI tools and the use of biometrics should be utilised for identity verification. Biometrics, e.g. verification methods like using fingerprints or voice prints of clients, helps to save time by avoiding manual identification verification 6. . 6
A key component of Enhanced Due Diligence (the due diligence which is required for high-risk clients) is a more comprehensive analysis of the Source of Funds and Source of Wealth of the client. Aside from collecting the correct documentation (e.g. salary slips, bank statements, inheritance wills), a thorough understanding of the ownership structure is also crucial so that any stakeholder who has played a role in the accumulation of the funds is known. It’s important to spot any so-called ‘shell companies’ 7 (companies which are registered only with an address, and which only hold passive investments) in the ownership structure, because these companies are often used as a tool to hide beneficial ownership information, to launder money or to commit tax fraud.
The use of adverse media screening tools in client due diligence is vital. These tools need to amalgamate information from all available online and offline sources, including negative news articles, sanctions and watchlists and whether the name is mentioned in publications like Panama or Paradise Papers (to check for links to fraud or tax evasion). 8 . Compliance departments are burdened by the frequent occurrences of “false positive” hits, i.e. hits which the system flags as being “true hits” (which relate to the real person) but after further analysis, prove to relate to a different person. These false hits can be abundant and all of them need further investigation to assess if/why they are immaterial. The more client data that can be inserted in a system during screenings, the less likely it is that false positive hits will occur.
AI could potentially be used for improving the accuracy of adverse media screening platforms. Generative Adversarial Networks (GANs) use two deep neural networks- a generator network and a discriminator network, which compete in an adversarial game against each other with the aim of being able to spot ‘fake data’ (false hits in these cases). The generator network produces the fake data, whilst the discriminator network tries to identify if the data is fake or real. The two networks compete against each other, and both update themselves if they aren’t efficient enough . 9 T.The logistics of using a GAN in this context would need to be tested by software engineers, but it’s a prospect that looks promising.
Natural Language Processing (NLPs) tools are strong assets that can be used for adverse media screening tools. NLPs, and ideally Multilingual NLPs are AI tools that can be used to rapidly analyse text, languages, as well as the contextual nuances within the text. NLPs can also extract key information from documents and classify them according to their relevance or the risk factors contained within, potentially minimising the need for manual input. However, NLPs still have their limitations and, as of the present, they still struggle with understanding context and where words and phrases have multiple meanings . 10 .
AI tools should be used as an asset and not for the sake of replacing human input which is also still crucial. AI tools can accommodate a transition towards a ‘Perpetual KYC’ model where transaction monitoring, identity verification and adverse media screenings can be conducted on an ongoing basis, rather than at specific intervals. This can both save time and increase the chances of financial crimes being detected . 11. AI tools can spot anomalies and changes in transactional patterns of clients which can be particularly useful when dealing with clients who work in industries where transactions are made frequently.
What is crucial in the fight against financial crime is for information to be exchanged between FIs on a cross-border basis, in a manner which is legally permissible. Tools which would allow this efficiently and without breaking GDPR rules are ‘Privacy Enhancing Technologies’ (PETs). PETs maximise the use of data by minimising the risks normally inherent to the processing of data by, for example, allowing privately held data to be used without disclosing copies of the data . 12 . Artificial Intelligence tools are a work in progress and still feature numerous limitations. The effective use of AI tools requires training the AI models with large amounts of high-quality training data, which will take time, money and effort for FIs, or in some cases, there can be a risk of regulatory breaches with the use of this data.
References
1 https://www.europol.europa.eu/crime-areas/economic-crime/money-laundering
2 https://smartkyc.com/utilising-ai-advanced-adverse-media-screening/
4 https://www.sas.com/en_be/home.html
5 https://www.tookitaki.com/products/anti-money-laundering-suite
6https://www.moodys.com/web/en/us/kyc/resources/insights/biometrics-in-customer-onboarding.html
7 https://financialcrimeacademy.org/unmasking-money-laundering-through-shell-companies/
8 https://offshoreleaks.icij.org/
9 https://aws.amazon.com/what-is/gan/
10 https://www.iso.org/artificial-intelligence/natural-language-processing#toc1