Screening for Sanctions risk in Trade Finance has always been a high risk, high pain game. But the vast majority of hits generated by standard list-based screening systems have no nexus to Sanctions at all. Now finally Trade Finance banks are not putting up with such inefficiency. New artificial intelligence (AI) techniques are helping banks dramatically reduce time spent on irrelevant Sanctions hits - and in fact eliminate those hits in the first place.
The leading global Trade Finance banks process millions of documents on behalf of clients each year. These documents contain even more millions of data points, including ports, goods descriptions, vessels, countries and so on - all of which and more need to be checked for potential OFAC and other Sanctions exposure. The cost of missing an actual Sanctioned entity or location is high. This is exacerbated by the duration of most Trade Finance and open account finance transactions, during which a previously non-Sanctioned party could appear on a Sanctions list.
The standard means of detecting a potential Sanctioned party to the transaction is to screen every field in every transaction against multiple Sanctions lists, using fuzzy matching logic. So that any name that looks like it may be on a list is flagged by the system for human review. The vast majority of the hits identified by the vast majority of banks, even operating in high risk locations and high risk sectors, are not related to Sanctioned entities at all.
Furthermore, because a potential Sanctioned entity, say the supplier of a bank's importing client, appears many times on many documents (invoice, bill of lading, insurance certificate and so on) multiple hits will be generated by bank screening systems. All of which need to be dispositioned manually.
Lists of Sanctioned and potentially Sanctioned people, corporations, vessels and so on have ballooned in recent years, partly due to Russia's invasion of Ukraine as well as geopolitical tension such as between the US and China over sensitive technology. This has worsened the problem any bank involved in Trade Finance faces when screening for Sanctions or export control risks.
For many years, banks in Trade Finance have accepted this is a reality: the cost of doing business, so to say. A lot of manual effort with a high potential for manual error in transcribing detail from documents into core trade systems. And many hits on many transactions using fuzzy logic. Often one transaction in three or more with multiple hits.
Methods to bring down the hit rate or improve the dispositioning process have been only moderately successful. These include "white-listing" where banks add counterparties (for example) they know to be non-client "good guys" who trade with their clients so that potential hits are suppressed. But the hit rate remains stubbornly high.
Replacing fuzzy matching logic with AI has changed the game. This includes screening based on natural language processing (NLP) which is dramatically more accurate than a fuzzy match. Measuring a name, an address or similar using NLP, together with field-to-field screening is just the start. If it is known that a certain data point represents a vessel name, this should be screened against Sanctioned vessels using intelligent algorithms. The same goes for corporations, ports, countries.
But let's not stop there. Generative AI has already proved that, despite some significant downsides, there is massive potential in large language models (LLMs). From a Sanctions screening perspective, the ability to scan vast amounts of publicly-available data to arrive at a more accurate decision on the potential for a party to potentially be a Sanctions risk is clear. While the downsides of Generative AI - such as the current inability to tell between a rule and a preference - are relatively irrelevant.
In a relative short space of time, Trade Finance banks who have long put up with high levels of inefficiency in Sanctions screening, which has occasionally led to compromised ability to actually detect Sanctions risk, are now radically changing their approach. This is for the better.