Skip to main content
New Frontiers in Using AI to Fight Fraud and Money Laundering

In our brave new world, the use of artificial intelligence can be a contentious issue. For many, the question of AI as friend or foe is not yet resolved. Those who work in fraud management provide a shining example of how AI can be a force for good, and have been doing so for over quarter of a century.

Despite being early adopters of AI, now is not the time for fraud and financial crime specialists to rest on their laurels — they are involved in an ever-escalating arms race with criminals who also use such technology to launch their attacks on financial institutions.

As Julie Conroy, director of Aite Group’s Fraud and AML practice, said at the Finovate conference last month, “While we’re meeting to discuss how to tackle fraud and financial crime, elsewhere the criminals are holding their own conferences to plan their attacks.” Conroy pointed out that fraud and money laundering are financing some of the worst crimes society faces, including human trafficking, terrorism and the operations of drug cartels. Those working in fraud and AML compliance at financial institutions are heroes and arming them with the most effective weapons not only protects their businesses but also society.

Artificial intelligence and machine learning to combat fraud and money laundering is a worthwhile investment. In a recent survey by research and analyst company Ovum, over 80% of banks who have invested in artificial intelligence believe that it generates return on investment.

So, given the consensus that AI and machine learning are an important weapon fight against financial crime, what developments will keep fraud and AML compliance functions on the winning side?

Fast to Deploy, Easy to Adapt

Fraud changes fast — as one avenue closes, fraudsters and money launderers look to the next opportunity. This can be seen with the launch of EMV (chip for card payments) in the USA, which forced the criminals to turn their attention to card not present (CNP) fraud. The shift of focus from EMV to CNP was a large-scale change, but smaller changes happen daily.

Financial institutions know they must respond quickly, and to do this they have invested in data science teams. There is no doubting the abilities of these data scientists to understand the issues they face and to build models that can work effectively – the issue they face is having the technology they need to deploy them.

Banks risk wasting their investment in data science teams if they don’t also provide them with the tools to operationalize the work they have done. As Doug Clare, who oversees FICO’s fraud and compliance solutions, said at the Finovate conference, “Banks need to pivot quickly on their experience of the financial crime they are seeing and get the models they develop into operation fast – but without investment in the right platforms they can’t do that. It’s very frustrating to have the answers but not to be able to deploy them in a timely manner.”

Financial Crime Is Fraud AND Money Laundering

For many years, financial institutions have operated separate functions for fraud and AML compliance – criminals do not work in such siloes and that gives them an advantage. Things are now changing, and as a recent Ovum survey shows, more than two-thirds of banks have strategic plans to integrate their fraud and AML functions – not surprising given that there is an up to 80% overlap in the requirements of both. By doing this, financial institutions look to benefit from better detection and operational cost savings.

They can also benefit from the cross-pollination of ideas and practices. In particular, AML compliance professionals can learn from their counterparts in fraud about the use of AI and machine learning. 38% of AML professionals report concern over the level of suspicious activity reports (SARs) that they are filing defensively, which significantly increases workload. Given that 42% also report that obtaining skilled and experienced staff is a major business issue, they need a new way to ensure that they accurately detect cases and do not spend time investigating and managing cases when it isn’t necessary.

In the past AML professionals have been concerned that the use of AI and machine learning to detect money laundering will not meet regulatory requirements. More recently, regulators including the Financial Conduct Authority in the UK, the U.S. Treasury Department’s anti-money laundering unit and federal banking regulators have indicated that they accept the use of AI and machine learning but that they encourage such innovative approaches.

Explainability – the Computer Can’t Just Say NO

As AI use grows, there is a concern that decisions that impact peoples’ lives are made by machines. Legislation such as the European Union’s General Data Protection Regulation (GDPR) says that individuals have the right not to be subject to a decision that has a legal or similar effect upon them and, that is based solely on automated decision-making (without human intervention).

This does not undermine the value of using AI to make decisions – what it does mean is that those decisions must be interpretable and that the factors that led to that decision must be identifiable, understandable and explainable. Organizations that deploy AI and machine learning to detect fraud and money laundering must therefore take care that the models they use are not ‘black box’.

AI models are in important tool in making decisions faster and more accurately, but they are not infallible. As FICO Chief Analytics Officer Scott Zoldi explains in his post ‘Bank of England Validates Need for Explainable AI’  the sheer size and complexity of these models make it difficult to explain their operating processes to people. Fortunately, all is not lost – the development of techniques outlined in Zoldi’s blog make it clear that it is possible to make artificial intelligence explainable, if you are using the right models.

Artificial intelligence and machine learning have been used by fraud management professionals for almost three decades and are now being adopted by their colleagues working to prevent money laundering. This does not make it old news; indeed, the trends identified in this article mean that it is a more exciting and innovative field to work in than ever before.

We will be discussing the latest innovations and best practices in AI and machine learning for fraud and money laundering at FICO World 2019 next month. Join us for the sessions, and to see Scott Zoldi discuss challenges with AI Now co-founder Kate Crawford in the closing general session.

related posts