Vladimir Putin recently endorsed artificial intelligence as the key to world domination, and hence it’s his new BFF. Elon Musk just admitted that SpaceX was really an escape pod to other planets for when AI goes full-Terminator. But in the meantime, AI may significantly reduce credit card fraud, before it figures out how to enslave humanity.
Automated fraud detection is loaded with drawbacks currently, particularly the high ratio of false positives they generate. The challenge lies in instantaneously parsing the near-infinite data points required to approve or decline each transaction. It’s an inherent problem with big data: you can have all the information in the world, but what to do with it all? MasterCard and Visa are currently in a bit of an arms race looking at ways AI can change this to make machine learning leverage big data more reliably.
Visa is reported to have tapped German and Israeli startup Fraugster (yes, with a g) to run a "self-learning algorithm" through their transaction volume to detect fraud in real-time. Fraugster claims they can reduce fraud by 70%, and increase conversion rates by 35%.
MasterCard calls their technology "Decision Intelligence" in what they claim is the "first use of AI being implemented on a global scale." Ajay Bhalla heads up the effort at Mastercard, which he characterizes as a more holistic approach. Rather than the algorithm scoring individual transactions, Decision Intelligence looks at data from the merchant, the device, and the cardholder’s history, pooling it all to score a transaction.
It’s hard to tell what differentiates these programs from one another since everyone is using the same vapid tech-speak to sell their innovations. But the proprietary nature of this competition raises some concerns. While one scheme might deliver on AI fraud prevention as promised, fraudsters will simply flock to any of the other weaker schemes to get around it. Just like old times. Also, will this AI treat all merchants fairly across the low-to-high risk spectrum?
However, one tangible benefit that this race to the smartest AI might bring is that it paint a more accurate picture of what fraud actually looks like. Pulling from the impossibly large universe of data points, perhaps machine learning can teach us how to better detect and fight fraud systemically. Or maybe AI can solve the problem by becoming self-aware, so we can have better problems to worry about, like survival.