Mastercard doubles compromised card detection speed and cuts false positives 200% with generative AI fraud technology

Payment card fraud operates at a scale that strains traditional detection systems. Fraudsters harvest millions of card numbers through spyware, malware, and physical skimming devices, then monetize stolen data by posting partial 16-digit card numbers on dark-web marketplaces — selling them to secondary criminals before banks can respond. Because only fragments of card data appear on illicit sites, conventional rule-based systems could not reliably reconstruct full card identities or predict exposure fast enough to intervene. For a network processing billions of transactions across millions of merchants globally, the lag between card compromise and bank notification created a window where fraud could propagate unchecked, resulting in financial losses and eroded cardholder trust across the ecosystem.

Mastercard developed a generative AI-based predictive technology that reconstructs complete 16-digit card numbers from the partial details posted on illegal websites — effectively closing the gap between data theft and bank notification. Built as an enhancement to the existing Cyber Secure platform (launched 2020), which already tracked cybersecurity risk profiles across the payment ecosystem, the new system applies large language models and generative AI to analyze transaction patterns across billions of cards and millions of merchants simultaneously. Rather than relying on known fraud signatures, the model identifies novel, complex fraud patterns by inferring relationships across the network. When a card is assessed as compromised, the system generates proactive alerts to issuing banks, enabling card blocking and reissuance before fraudulent transactions can occur — shifting Mastercard's posture from reactive to predictive.

The generative AI upgrade delivered measurable improvements across detection, accuracy, and merchant risk assessment:

• 2x increase in the rate at which potentially compromised cards are identified
• 200% reduction in false positives during fraudulent transaction detection, cutting unnecessary friction for legitimate cardholders and reducing alert fatigue at issuing banks
• 300% faster identification of at-risk or compromised merchants across the network

Beyond the metrics, banks now receive faster and more precise alerts, enabling quicker card-blocking and reissuance cycles. The reduction in false positives is operationally significant: fewer erroneous flags mean bank fraud teams can focus resources on genuine threats rather than triaging noise.

• Generative AI's ability to reconstruct complete data from incomplete inputs is particularly valuable in fraud contexts, where attackers deliberately fragment stolen information.
• Network-level scanning across billions of cards surfaces emerging fraud patterns that individual issuers — seeing only their own data — cannot detect in isolation.
• Reducing false positives is as strategically important as increasing detection rate; high false-positive volumes erode bank confidence in alert systems and harm legitimate cardholders.
• Layering new AI capabilities onto an established platform (Cyber Secure) allowed Mastercard to compound existing ecosystem trust rather than rebuilding from scratch.
• Proactive card blocking, triggered by predictive signals rather than confirmed fraud, fundamentally changes the cost equation for card-present and card-not-present fraud.