Mastercard keeping humans in the loop of AI fraud and risk-related decisions
While artificial intelligence (AI), machine learning (ML), and automated machine-driven processes are increasingly important in providing better cybersecurity, as well as fraud and risk management, in the financial services sector, Mastercard believes there will still be a place to keep a human in the loop.
“We do believe that humans will continue to play an integral role,” Mastercard APAC executive vice president and head of services Matthew Driver told ZDNet.
“As we increase the number of areas where we apply tools, there is a need for human oversight and reviews in many stages but critically in system design and control systems.”
Rather than being mutually exclusive, Driver said Mastercard sees the roles of humans and the application of automated tools to be complementary.
“Humans are able to make manual reviews and, with experience, can help move these decisions to rules or embed them into models. But machines cannot attribute or deduct causality, so while there will always be newer areas where we are applying AI and modelling, there is a constant need for these to have a human overlay in design and governance,” he said.
Driver told ZDNet that Mastercard is actively embedding AI throughout the organisation, touting “significant” results in terms of both accuracy and operational efficiency.
“We have AI applications running across many disparate functions including … improving the quality of our financial forecasting globally, natural language processing systems to help in areas like developmental training or talent management, anomaly detection in expense reporting for audit purposes, and removing the need for repetitive tasks with robotics process automation tools,” he explained.
According to Driver, these applications are raising the precision by which Mastercard manages its own business. Specifically, it has improved the speed and accuracy of applying AI and ML to large datasets and mundane tasks. But keeping the human involved has meant such decisions are confirmed before they are released into the wild.
He said Mastercard has been using AI capabilities for almost 10 years, saying the tech is built into the core of its network.
“Since 2018 we have helped banks save $14 billion globally by using AI to prevent criminal attacks,” he said. “Furthermore, the fraud tools we have created are becoming increasingly sophisticated as they learn how criminals are evolving their tactics.”
Touching on the payments ecosystem as a whole, Driver said it is still experiencing a rise in ransomware, with small businesses at just as much risk as the bigger end of town.
“The vast majority don’t have a cyber team, cyber specialist, or an IT manager. For most small businesses, cards and payments are a critical lifeline, especially in an increasingly connected digital economy, and most do not have sufficient access to security expertise to ensure they are adequately protected,” he said.
He also said there remains a need for constant customer education as fraud today is becoming increasingly sophisticated.
What are customer insults, and what does machine learning have to do with it?
HITL is a mix and match approach that may help make ML both more efficient and approachable.
The US and the UK are the most common sources of stolen credit card data.
The non-profit organisation that will be based in Switzerland aims to protect victims against cyberattacks and assist them in the aftermath.
Captured through malware and sold on the Dark Web, the “digital fingerprints” of your web browsing can be used to impersonate your identity online, as described in a new report from IntSights.