AI accelerates AML processes across financial services

Financial regulators across Europe continue to levy steep enforcement fines against banks for failures to comply with know-your-customer (KYC) and anti-money laundering (AML) regulations. At the end of 2021, the Financial Conduct Authority (FCA) fined two of the UK's largest banks, HSBC and NatWest, a total of £328.95 million ($436.1 million) for failings in their money laundering processes.

Meanwhile, members of the European Parliament are calling for cryptocurrencies to be governed by the European Commission's Anti-Money Laundering Authority, as illicit organisations continue to find new methods for laundering money through the financial system.

Money laundering is a process that criminals use to hide the illegal source of their funds. By passing money through multiple, sometimes complex, transfers and transactions, the money is "cleaned" of its illegitimate origin and made to appear as legitimate business profits.

Technological advances in areas such as digital banking, online account opening, open banking and cryptocurrency have made tracking the source of funds and uncovering suspect patterns and behaviours far more resource-intensive for financial institutions and their regulators. Traditional methods of automation are simply unable to keep up with the increasingly sophisticated ways in which criminal organisations abuse the financial system, or with the rapid evolution of technology.

Artificial intelligence (AI) is therefore one of the most promising AML tools available to bankers and regulators. It can be thought of as the development of computer systems that can perform tasks autonomously, ingesting and analysing enormous volumes of data and then recognising patterns in that data.

AI tools are focusing primarily on the development of systems which can perform tasks that would otherwise require human intelligence to complete, and at speeds beyond any individual or group's capabilities. Financial institutions use AI across their businesses to power applications ranging from risk management for capital markets to virtual assistants for customer support in consumer finance.

Fraud prevention is another priority use case for AI in financial services. In fact, 2022 research carried out by NVIDIA shows that two of the top five AI use cases for investment are "Fraud detection: transactions and payments" and "Fraud detection: AML and KYC".

Why is AI such an effective AML tool?

First, AI models and algorithms can consume and synthesise massive volumes of data. These inputs are not limited to traditional types of tabular data (i.e., transaction ledgers) but can also include unstructured data (i.e., audio, video and geospatial inputs). Furthermore, AI can ingest the data and act on it in near-real time, enabling authorities to stay in step with the movements of bad actors rather than remaining days or weeks behind.

AI models are designed to detect anomalies in the patterns of data they are ingesting by scoring those behaviours relative to expected benchmarks, so that banking compliance officers are alerted when potentially nefarious interactions may occur. The investigations tied to these alerts are often led by compliance personnel within banks, and are therefore time-consuming and costly.

Traditional rules-based methods — a common technique from before the advent of modern AI — have a high false positive rate, meaning investigators' valuable time is wasted on the wrong transactions.

Knowledge graphs to transform fraud detection

Leading banks are therefore employing AI deep learning techniques such as GANs (generative adversarial networks) and GNNs (graph neural networks). Given enough historical financial transaction data, deep learning-based approaches are better at pattern matching than rule-based approaches, as they can generalise to learn fraud schemes and then use that AI to identify active fraud schemes in the data.

As an example, GANs can generalise from training data to identify patterns in transactions that are indicative of money laundering. That is, having been shown some patterns in real situations, the corresponding deep neural networks (DNNs) can generalise from the examples to identify similar and modified patterns that could get around the static rules, but are similar enough to the old pattern that they are caught by the DNN. This makes it harder for criminals to avoid detection. They will no longer be able to make small adjustments to the way in which they launder their money to get around a relatively static set of rules.

In addition to GANs, GNNs are another DNN technique that allows investigators to evaluate relationships between any number of parties to flag potential money laundering behaviour. The concept is to construct a heterogeneous graph from tabular data and train a GNN model to detect suspicious transactions and complex laundering activities, as criminals work collaboratively in groups to hide their abnormal features but leave some traces of relationships.

The relationships identified by GNN-based models is vital, as the AI can identify previously unidentified relationships across entities. With the advantages of capturing relations, GNNs are more capable of detecting collaborative laundering activities than traditional models.

AI's positive impact on one bank's AML operations was proved through a recent collaboration between Swedbank, Hopsworks and NVIDIA. In this example, Swedbank and Hopsworks trained GANs as part of the bank's fraud and money laundering prevention strategy. Using this solution, Swedbank was able to reduce its false positives by 99% compared with existing rule-based systems, and to create an estimated increase in efficiency for investigators (time to investigate) of more than 50% within five years.

In addition to leveraging AI for intra-company data ingestion and analysis, federated learning techniques will enable improved data sharing across departments, jurisdictions and companies because of its ability to maintain compliance with data sovereignty and privacy regulations.

The greater volume of data available for analysis by AI models will significantly improve the accuracy of the models and will make it even harder for bad actors to successfully launder money. Furthermore, AI technology such as robotic process automation and optical character recognition will aid investigators' analysis of documents, creating further efficiencies and reducing error rates across the process.

Originally published by Thomson Reuters © Thomson Reuters.