Unmasking money-laundering gangs with AI and graph databases
"Criminals have become adept at creating intricate networks of identities, shell companies, and banking accounts to obscure their activities."
Money laundering remains a high-stakes cat-and-mouse game, with the UN estimating that between 2% and 5% (603 billion to 1.75 trillion pounds) of the world’s GDP is laundered each year. In the UK alone, financially motivated offences account for over a quarter of all crime.
As the UK government considers folding the Payment Systems Regulator (PSR) into the Financial Conduct Authority (FCA) to reduce short-term costs, the burden on resources to detect illegal financial activity will likely intensify. In parallel, the European Union’s Digital Operational Resilience Act (DORA) drives a tougher stance on digital operational defences across the financial sector, compelling institutions to lock down loopholes the criminals would otherwise exploit.
This shift comes at a time when payment scams and payment methods are evolving rapidly, stoking concerns about money laundering and financial wrongdoing. Although the PSR and FCA have overlapped in some functions since the PSR’s inception, some observers question whether blending these regulatory roles will be effective enough to stop the ever more cunning criminal networks. Adding to the pressure, in January the FCA published an updated analysis on money laundering through the markets, highlighting the risk of capital markets being misused to move illicit funds under the guise of legitimate transactions.
In light of these complexities, financial institutions are increasingly turning to emerging solutions such as AI and graph databases, to investigate suspicious behavior more deeply. With money laundering tactics evolving, it’s crucial for the industry to adopt fresh approaches that can keep pace.
Why money laundering slips under the radar
Money laundering is notoriously difficult to detect because illicit transactions often blend seamlessly within everyday banking activities. With so much at stake, criminals have become adept at creating intricate networks of identities, shell companies, and banking accounts to obscure their activities. These multiple layers can look entirely innocuous on the surface, which is precisely the problem for traditional anti-money laundering (AML) systems.
Conventional AML measures typically look to identify deviations from standard patterns within discrete data and transactions. At a technical level, they are based on a relational database model where data is stored in rigid tables and columns. The underlying assumption is that ‘normal’ activity can be measured against an outlier, yet if a criminal has always operated illegally, their behaviour might not register as unusual.
This approach is often ill-equipped to handle vast data sets of financial records, and it struggles to pinpoint the patterns that might indicate a hidden network of illegal transactions. The result is a deluge of false positives, with investigators spending too much time on fruitless leads. Meanwhile, genuine threats can slip past undetected if the transactions in question appear superficially consistent with the criminal’s established history.
AI and Graph Databases: new tools for an old issue
A shortcoming of many relational database-oriented AML solutions is their inability to incorporate bigger-picture insights. What is needed is the ability to follow a trail from one account to another; a 360-degree view of complex money laundering networks is necessary to flag connections between assets and individuals - and is something that knowledge graphs can provide.
Financial institutions often have significant blind spots regarding transactional fraud, because criminals spread their activities across various accounts or even different financial providers. A launderer rarely sends money directly from one bank to another in a linear path; rather, they route through a sophisticated web of ‘mule’ accounts. This makes it difficult for any single institution to gain full visibility into the end-to-end flow of money.
Graph database technology, in particular, is well-suited for AML efforts, as any number of qualitative or quantitative properties can be assigned to data, describing complex patterns coherently and descriptively. Graph databases use individual data such as ‘person,’ ‘account,’ ‘company,’ and ‘address,’ along with their connections to one another, such as ‘registered at’ or ‘transacted with,’ to uncover complex connections. Specifically, looking for key individual data and connections in this way means financial institutions can uncover intricate networks quickly, flagging suspicious connections that would otherwise remain hidden.
A notable example of this is how the International Consortium of Investigative Journalists (ICIJ) used Neo4j to sift through millions of leaked files in the Panama Papers investigation, quickly making out hidden offshore structures and exposing cross-border links between individuals and assets. This same approach gives today’s financial institutions a powerful edge in spotting suspicious connections that might otherwise remain buried.
Cutting-edge tech vs criminal networks
Graph database software and AI-driven analysis are taking AML investigations to a new level of detail. By finding patterns in real time, these solutions help investigators stay one step ahead of the criminal networks and their dirty money. Equipped with graph database technology and AI, financial institutions can track hidden transactions and unravel money-laundering plots.
The FCA’s emphasis on continuously refining controls and systems, as noted in its latest analysis, lines up with this technological advance. Both public bodies and private firms alike must keep pushing innovation to tackle the evolving threat of money laundering, and the government’s plan to merge the PSR with the FCA underscores the desire for a unified regulatory front, aimed at thwarting criminals who thrive on oversights and gaps.
By embracing graph databases and AI, the banking and finance sector is better equipped to detect and disrupt unlawful schemes before they grow too large to handle. As these solutions become more integral to AML workflows, money launderers will discover that simply hopping from one account to another is no longer a dependable way to remain invisible. The end result is a more robust and transparent financial system, exactly what regulators and legitimate market participants are working to achieve.
Michael Down is Head of Financial Services Technology at Neo4j
