About
Preventative AML Models
Innovation presentation
The money incoming of terrorism and other criminal activities like human-drug traffic, arms smuggling etc., laundered in system through usurping the rights of innocent people reached almost $2 trillion. Could the money-laundering be prevented for humanity and the reliability of financial sector?
Definitely, instead of generating thousands of alerts with hundreds of traditional rule-based scenarios and examining all alerts by the compliance teams, AI models newly developed will be the solution for AML future. The scenarios cannot rapidly adapt to the new money-laundering ways. In this context, a set of AI models pioneering the sector have been developed in 2020 first time ever by ING Turkey for detecting potential criminals in the future, clarifying complex transactions, automatically closing alerts.
Uniqueness of the project
We think that the rules of the game have been written again with the models presented using the Customer Activity Monitoring Team in Compliance Group (CAM).
• The money laundering transactions are able to be detected with the Probability of Suspicious (PS) Model approximately 90 days before a suspicious transaction happens.
• The Smurfing transactions widely used as a money-laundering method are unable to be identified by traditional detection applications; however, our model is able to identify this method around 85%.
• We are able to automatically close the no-risk retail alerts around 30% and determine what alerts should be focused on instead of monitoring the millions of transactions with the rule-based scenarios and investigating thousands of alerts. In conclusion, it has been provided 20% out of the workforce in the CAM and prioritized risky transactions.
The many money laundering attempts valued millions of dollars have been identified and avoided already through our models; moreover, it helped to state the criminals to the competent judicial authority. In this approach, the efficiency in the CAM is increase besides that our institution is prevented to be fined millions of dollars by regulators. The safety of our institute and the world has been indirectly provided.
Furthermore, all AI models have been presented to an independent validation unit to prove the model's performance and correctness. As a result of the validation reports, all models are found as powerful and approved to be used.
