Anti-money laundering (AML) Suspicious Activity Reports (SARs) are of poor quality due to the reliance on manual processes, according to Martina King, chief executive officer of Featurespace.
“I think the problem is that so many of the systems and processes around AML are manual. The promise of machine learning is that you should be able to rely on the machine to support the decision making, and with the sheer volume of digital events that are taking place in any large organization’s data set it almost becomes unwieldy to just have a human process to this problem,” says King.
“With the systems that currently exist the challenge that we’ve been told for quite some time is that they just have too many false alerts. That was exactly the same problem when we first entered into the fraud market,” she says.
On June 18, the UK’s Law Commission presented a parliamentary report on SARs used by banks and businesses to inform the United Kingdom Financial Intelligence Unit (UKFIU) of suspected money laundering.
A large number of SARs are of low quality and contain limited, or even no, useful information, which is particularly worrying considering the UKFIU received a record number of reports for 2018, according to the findings.
“Money laundering is a blight on the UK’s economy and damages our international reputation. We must have a regime in place that allows law enforcement agencies to investigate and disrupt money laundering at an early stage,” said Professor David Ormerod QC, Criminal Law Commissioner at the Law Commission in a statement.
“But the reporting scheme isn’t working as well as it should. Enforcement agencies are struggling with a significant number of low-quality reports and criminals could be slipping through the net,” he said.
Earlier in June new chair of the European Banking Authority (EBA) José Manuel Campa told the FT that the authority’s supervision powers do not go far enough to stop Europe-wide AML.
“I don’t think the mandate that the EBA has received is the mandate that will solve that problem,” he said. “It is not a mandate to harmonize [AML], either regulation or practices, across the union. Because to start that process you first need legislation,” he said.
For King, as application fraud is on the rise, the industry is crying out for a method of predicting future fraud and moving away from manual processes.
“The more protection that you try to put into the customer journey, the more friction it creates and so the more abandonment occurs and obviously there is a loss on the back of that as well. I think the final part is once you close one door or attempt to close a door on a fraudster – where will they go next? That is where we can see application fraud for instance is an area that has been increasing, particularly in the UK,” says King.
“The industry has encouraged us to see if we can solve the AML problem in an optimal way. What I mean by that is, is it possible to create a methodology for identifying suspicious activity, prioritizing it to make it easier for an analyst, so their workflow is prioritized alerts? On top of that, is it possible to identify through these gigantic data sets – a way of finding suspicious activity weeks or months ahead of when the manual reviews would then get to those alerts?”
As firms are under much more pressure to manage their data capabilities, organizations such as Featurespace are now offering services to add intelligence and direction to large datasets.
“Now we can see in a large data set of alerts that we can reduce the overall volume of reports. Then to find the suspicious activity analysts only have to work with the top 5% of the rule-based alerts in order to be able to get 100% match of the suspicious activity reports – which is transformational for the companies that are trying to manage and monitor AML,” says King.
“In an era where doing only the basics is no longer enough, we have our anomaly detection which enables us to identify suspicious activity weeks and months ahead of the manual reviews taking place.”