AI’s fight against money laundering in trade

Blog | 7 June 2017

149682728265721.jpg

Mallinath Sengupta looks at the benefits artificial intelligence software could provide financial institutions in detecting and stopping money laundering in trade

International trade, which was valued by the WTO at US$16trn in 2015, is a tremendous engine of global economic growth, but it is also a magnet for criminal activity, especially money laundering. This places a significant regulatory burden on financial institutions, which are vital intermediaries in international trade, to identify and report suspicious activity that may indicate the movement of illicit funds, narcotics trafficking, terrorist financing, or other criminal activities.

The traditional, largely manual methods of identifying trade based money laundering (TBML) violations are laborious, costly and error-prone. Artificial intelligence (AI) systems offer a more rigorous and cost-efficient solution to protect financial institutions against the risks of non-compliance with TBML reporting requirements.

Scale and scope of the TBML problem 

The US treasury department’s Financial Crimes Enforcement Network (FinCEN), advised financial institutions that “criminal organisations use the international trade system to transfer value across international borders and disguise the illicit origins of criminal proceeds”. The level of money laundering accomplished through international trade is nearly impossible to measure accurately, given the illegal nature of the transactions, but it is agreed the amount is vast.

A 2009 International Narcotics Control Strategy Report (INCSR) estimated that the annual US dollar amount laundered via trade ranges into the hundreds of billions. The IMF calculated in 2011 that money laundering comprised approximately 2-5% of global GDP, which would currently amount to some US$3-5trn per annum. Additionally, U.S. Immigration and Customs Enforcement (ICE) reports that their trade based money laundering case initiations have increased since 2004.

Looking out for red flags

FinCEN has issued an extensive list of red flags that could alert financial institutions to possible money laundering, which include, but are not limited to, the following:

●     Payments made by an intermediary apparently unrelated to the seller or purchaser of goods, which may obscure the origin of the funds;

●     Amended LCs without reasonable justification;

●     A customer’s inability to produce appropriate documentation, such as invoices;

●     Significant discrepancies between the description of goods across different documents (bill of lading, invoice, packing list, and so on);

●     International wire transfers into US bank accounts received as payment for goods, especially where the ordering party does not reside in the country from which the wire originated;

●     Funds transferred into US accounts that are subsequently transferred out of the accounts in nearly the same amounts; and`

●     Foreign visitors opening multiple US bank accounts.

Business challenges in identifying TBML

Regrettably, the opaque nature of international trade finance makes it difficult to consistently identify the red flags, thus leaving the door open to TBML. As supply chains have become globalised, importers and exporters may not know one another personally. They rely on banks and other intermediaries to settle financial obligations. The importer establishes a letter of credit (LC) in favour of an exporter. Banking institutions typically draw against the LC to pay the exporters based on documentation that goods have shipped, rather than on physical inspection. And, compliance monitoring tends to be ‘low tech’ – with manual reviews of LC applications, exporter information, contracts, shipping documents (often poor-quality paper copies instead of digital versions), and other parts of the information trail.   

In addition, FinCEN has cautioned financial institutions to be alert for, “misrepresentation of price, quantity, or quality of merchandise”, particularly in conjunction with shipments of high dollar value goods. In practice, however, this can be hard to achieve in the complex universe of international trade. For example, a bank may not be able to detect under-invoicing or over-invoicing, since reference or benchmark pricing is not always readily available for industrial items (unlike commodities or consumer goods).

Among other challenges, financial institutions need to examine the relationship between party and counterparty (for example sister companies). Another area of exposure is validation of vessel information and route, as a vessel could be diverted to a port not on the original route list. There is also a concern as to whether goods being transacted are dual-use or not – that is, do products have a possible military use that could be exploited by criminals or terrorists. In terms of documentation, there are numerous cases of duplicate invoices being submitted for payment of the same goods.

Manual processes – obstacles to deterring TBML

Today’s processes are not up to the challenge of identifying and reporting suspicious activity that may indicate TBML. Current operations processes typically involve manual tasks and workflows with no automation. According to a survey of financial institution compliance executives conducted by NextAngles, 61% of respondents do not currently use an automated TBML system – and more than one in four don’t know if their institution has such a system.

The absence of rule-based alert generation systems has resulted in the manual creation of alerts, which is not only time-consuming and costly, but can lead to potential risk exposure as suspicious activity may be overlooked due to human error. It is also difficult with a manual approach to consolidate data from various systems within and outside a financial institution. Additionally, manual processes are challenged to keep up with constantly evolving AML and trade finance regulations and guidance.

AI technology provides solutions

Artificial intelligence (AI) systems are emerging as new weapons in the fight against trade based money laundering. In TBML applications, AI technology allows for the creation of rules-based ‘learning systems’ that can replicate the real world of international trade finance and become more expert over time. AI systems are able to perform tasks that normally require human intelligence, such as data gathering, pattern recognition, and discrepancy identification.For example, an AI-powered solution can search across multiple bank systems and business units to assemble the data to detect possible TBML violations, and also provide an appropriate audit trail.

An effective AI solution for TBML detection should be based on a knowledge model that ‘understands’ how a particular business ecosystem is supposed to work. That is, the system should recognise the types of goods, monetary amounts, countries of origin, and other factors that are relevant to transactions between specific types of buyers and sellers. For example, most businesses have a North American Industrial Classification System (NAICS) code that identifies the type of business. An AI solution could include a ‘negative list’ based on NAICS codes, which recognises certain types of companies that would not typically be doing business together. Through this and other means, the system could then determine whether an activity is within normal parameters or if it raises a red flag.

AI-enabled technology also offers such capabilities as:

●     Automatically spotting inconsistencies between trade documents, such as invoices, bills of lading, letters of credit, certificates of origin, and other transaction documents, thus saving investigation time;

●     Discovering hidden elements in a transaction that investigators might otherwise miss;

●     Generating alerts based on rule violations, and applying previously used investigation steps;

●     Iterative learning of the investigative processes, so that investigations can adapt and new steps can be learned as an improvement;

●     Semantically enabling the linking of relevant data; and

●     Providing dashboards that aggregate statistics on productivity, as well as indicate which red flags are leading to actual TBML violations.

Trade based money laundering is a pervasive, multi-billion dollar crime globally. But with AI technology, financial institutions can improve the accuracy, productivity and cost-effectiveness of their TBML solutions – while reducing the risk of non-compliance with increasingly strict anti-money laundering regulations.

Mallinath Sengupta is chief executive of NextAngles

Give Feedback