Mastercard Decision Intelligence and Transaction Fraud Monitoring

par Mastercard

AI Application

Faible confiance

Mastercard's AI/ML transaction fraud and risk scoring: Decision Intelligence ('a transaction risk monitoring solution powered by AI that helps you prevent fraud and approve genuine transactions in real time') and Transaction Fraud Monitoring (advanced AI and machine learning used by organizations to fight fraud). It works on payment-network data (card number, merchant, amount, time), fraud scores, IP and location data, and in some online products behavioral biometrics (keystroke timing, accelerometer, scroll position, mouse location), which the US privacy addendum treats as biometric and sensitive information. Mastercard says AI-driven security improvements raised fraud detection rates by 20% and up to 300% (vendor-reported). Rated high because the scores can influence whether individuals' payments are approved or declined, at network scale, using behavioral biometrics, and because the Global Privacy Notice's commitment against solely automated decisions with significant effects has an exception for fraud prevention. No published accuracy, false-positive or bias metrics and no model documentation were found in the pages read. The privacy notice says Mastercard generally acts as processor for financial institutions, so who makes and explains a final decline decision (issuer or Mastercard) varies and was not documented. I read only the descriptions on the AI and Agent Pay pages, not the product pages for these solutions.