Mastercard keeps tabs on fraud with new foundation model
Mastercard announced on April 3, 2024, a new large tabular model (LTM) designed to spot fraud in digital payments. The model, trained on billions of card transactions, will augment its existing security services.
The move follows a surge in online fraud and regulatory pressure for stronger transaction monitoring. Mastercard has long used machine learning for risk assessment, but this foundation model marks a shift to a dedicated, scalable architecture.
By training on structured transaction data, the LTM can capture nuanced patterns that text‑based models miss, improving false‑positive rates. It signals a broader industry trend toward domain‑specific foundation models, potentially lowering costs for merchants. However, the reliance on proprietary data may raise privacy concerns and limit cross‑vendor interoperability.
Merchants using Mastercard’s payment gateway will see tighter fraud detection, potentially reducing charge‑back costs. The model’s rollout will also pressure competitors like Visa and PayPal to accelerate similar initiatives. Regulators will monitor data handling practices closely.
- Mastercard launches LTM to improve fraud detection.
- Domain‑specific models may become industry standard.
- Privacy and cross‑vendor data sharing remain challenges.