Eve Whittaker, Market Planning Director at LexisNexis Risk Solutions, notes that the point is to ensure that anyone facing such sanctions has restricted access to financial systems.

“There are a number of official sanctions lists, drawn up by national governments, as well as by international entities, such as the European Union and the United Nations. These entities issue sanctions and restrictions directed at states, individuals, high-risk groups or legal entities suspected of being involved in illicit activities,” she says.

“Screening against sanctions lists fall under the category of financial crime compliance or AML regulations. From the perspective of financial institutions, this means they have a legal obligation to perform the necessary checks, and failure to adhere to these may be met by high financial penalties, as well as the attendant reputational risk.”

Whittaker says the challenge lies in the fact that traditional methods of screening simply cannot keep pace with modern business demands. Trying to manually process paper instruments, particularly in light of the sheer volume of transactions today, creates limitless potential for error. After all, data extraction and analysis, screening alert review and record keeping are, by their very nature, complex.

“Digitalisation is therefore critical. Generally, there are multiple lists to screen against. Compiling these together, ensuring they are always up to date, and then screening and checking counter-party names and transactions across all the lists very rapidly, while it also does this on an ongoing basis, requires automated systems to maintain an efficient and effective process.

“Artificial intelligence (AI) is particularly effective at eliminating false positives, which occur due to issues around how unstructured the data is, and also because a threshold is needed around what we call ‘fuzzy matching’. This is the leeway required, due to the fact that sometimes the precise spelling of names might not match, so there are parameters to ensure that nothing is missed, but this can lead to false positives.”

She adds that without AI, increasing payment volumes would mean deploying an ever-larger number of personnel to separate true and false positives on the alerts raised by matching algorithms.

“The next step lies with generative AI (GenAI), which excels at working with unstructured data like this. GenAI can discard alphabetically similar but semantically different false positives. This capability will drastically improve the effectiveness of screening, as well as making the effort of managing screening lists simpler and less time-consuming,” notes Whittaker.

“We are also seeing some interesting developments with regards to large language models, which can help to apply more structure to an unstructured data system. These models allow for more effective and accurate matching and help to improve the efficiency and accuracy of the process.”

She adds that with AI, there are numerous other ways for it to be applied, including implementing it in the pre-processing state – to organise and structure data – as well as to derive better quality matches. Furthermore, in the post-processing stage, it can assist in deciding which identified potential matches need to be prioritised.

“Ultimately, AI is a really powerful tool, with enormous potential to drive the efficiency and accuracy of screenings. I believe that if banks apply this properly, we will see a massive improvement in the quality of their sanctions screening programmes,” concludes Whittaker.