By Maarten Ligthart, Head of KYC CoE / Monitoring & Screening, ING
Interview ahead of the 3rd Annual Fraud and Financial Crime Europe Summit, taking place 1-2 April in London
What, for you, are the benefits of attending a conference like the ‘Fraud and Financial Crime Europe’ and what can attendees expect to learn from your session?
It contributes to my network, having the ability to improve my knowledge and be familiar with latest threats, risks and innovation.
In terms of what they can learn, that depends. I hope they can learn from my experiences on this topic and adopt in their own practice.
How can risk professionals utilise current AI and machine learning technologies, to mitigate the risk of financial crime?
A couple of elements are important to take into account: first of all it is all about data availability and data extraction. After that, the more technical part is crucial. So, you need to have in place a multidisciplinary team. Our content expertise is more on setting the attributes for anomaly detection, pattern recognition. It is important that you have strong supporting hardware, i.e processing power. It sounds silly, but we are dealing with huge number of data. The data needs to be restructured. A long story, short, it requires a lot of detailed prep work. Furthermore, it is all about exploration of your model, and learn from it.
Why does AI and machine learning play such a key role in the future of financial crime prevention?
More general: AI and advanced analytics (AA) can help understand patterns in behaviour that are impossible to detect by human beings and the perpetrators are not even aware of themselves. by leveraging known cases we can train machine learning models to move away from the rule-based approach of transaction monitoring and let the machine generate new rules
As said above, the human power is lacking, we can’t handle the huge number of data and don’t have the intellectual power to analyse all patterns, first define it and secondly analyse it that you really have a potential transaction or subject to investigate further.
Bear in mind: crooks are using machine learning as well so banks also need to, but also be realistic and ai is an enhancement, not a replacement of skilled people.
What are the challenges of using AI and machine learning technologies with the latest compliance and GDPR laws?
And how can it be overcome? GDPR laws are used as a generic obstacle. If you are able to point exactly what you want, after that you can explore this further. For some countries, you need local tailored solutions. So, the GDPR is not the main challenge. The main challenge is the ability of using the data for analytical purposes.
GDPR makes it difficult to use all data. Also transparency as the more powerful machine learning models make a lot of decisions and consume a lot of data in order to be effective. regulators demand that you understand these models. Banks should invest in educating and permanently train employees. Note that GDPR can somehow be resolved with encryption.
Currently what is the biggest issue within financial crime that AI and machine learning technologies will one day overcome?
How do you think it will overcome this issue? We need to convince all relevant stakeholders that we are able to explain the so called ‘black box’, i.e. the audit trail on generating algorithms. At this moment, most of the regulators are struggling with this subject as this is challenging. The slogan: “if it is not documented, it does not exist”, seems to be valid in this regard.
