Mastercard Unveils AI Engine for Smarter and Safer Payments

Mastercard has developed a large-scale AI model that could transform how the payments giant detects fraud, enhances loyalty programmes and delivers personalised experiences to its users.

The foundation model, powered by NVIDIA and Databricks technology, is one of the first payments-specific AI systems designed to understand the complexities of global commerce through billions of anonymised transactions.

The credit card company is leveraging NVIDIA NeMo AutoModel and NVIDIA accelerated computing alongside the Databricks platform to develop its proprietary transaction foundation model.

This system is trained on hundreds of millions of transactions and early results suggest it could outperform advanced machine learning techniques across various operations.

For fintech companies watching developments in AI-driven payments infrastructure, Mastercard's approach offers a glimpse into how transaction data can be harnessed at scale to create more intelligent financial services.

Steve Flinter, Distinguished Engineer at Mastercard, explains that the foundation model represents a different approach to deep learning.

"Our new foundation model is a different kind of deep learning neural network, called a large tabular model, or LTM, which is trained on structured data, such as large-scale tables or datasets," Steve says.

The company plans to expand the model's training to include additional payment transactions and diverse datasets encompassing merchant location, fraud patterns, authorisation data and chargeback information.

"As we train the model on more data and more kinds of data, it will be able to provide more insights and predict future transactions with greater accuracy," Steve adds.

Enhanced fraud detection capabilities

Mastercard's immediate priority centres on deploying the model to strengthen its cybersecurity infrastructure.

The large tabular model learns key characteristics and patterns with minimal human intervention, marking a departure from traditional security approaches.

Existing security models typically require data scientists to manually engineer features from raw transaction data to identify anomalies such as unexpected spending spikes.

This manual process is time-consuming and relies heavily on human expertise to define what constitutes suspicious behaviour.

This new system analyses data independently, uncovering connections that human analysts might overlook.

The AI-driven approach allows the model to identify subtle patterns across millions of transactions that would be impossible for humans to detect manually.

The improved analytical capability has already demonstrated potential in reducing false positives, particularly by better recognising legitimate but infrequent high-value purchases such as engagement rings, which current models often incorrectly flag as fraudulent transactions.

This improvement could significantly enhance the customer experience by reducing unnecessary payment declines.

Applications beyond security

The large tabular model's potential extends far beyond cybersecurity applications.

Mastercard could deploy the technology to enhance rewards programmes, refining how the company identifies relevant offers and incentives for individual cardholders based on their transaction history.

The system could also be used to refine personalisation models, enabling more accurate predictions of customer preferences and behaviour.

This capability would allow Mastercard to deliver more relevant services and communications to its users.

Additionally, the technology could optimise portfolio management and improve data analytics tools across the company's operations.

The system's flexibility could also enable the firm to consolidate the thousands of AI models it currently maintains for different markets, use cases and customer segments.

For fintech businesses, this consolidation approach could signal a shift towards more unified AI architectures that reduce operational complexity while improving performance.

The ability to serve multiple functions from a single foundation model could lower costs and accelerate deployment timelines for new features and services.

Broader fintech industry adoption

Other financial services firms are achieving results with comparable technology.

Revolut built a transaction foundation model using masked prediction, a self-learning methodology, to improve fraud detection and accurately forecast customer purchasing behaviour.

The digital bank utilised NVIDIA's AI stack, including NVIDIA Hopper GPUs, the NVIDIA cuDF library and the NVIDIA Nemotron family of open models.

This technology infrastructure enabled Revolut to process vast amounts of transaction data efficiently.

According to Revolut, this implementation resulted in a 20% increase in fraud detection precision, enhanced credit risk predictions and a 9.6% improvement in cross-sell accuracy.

These metrics demonstrate tangible business value from deploying foundation models in financial services.

These outcomes demonstrate how foundation models trained on transaction data could reshape core fintech operations from risk management to revenue generation.

The technology represents a fundamental shift in how financial institutions approach data analysis and decision-making.

Future development plans

Mastercard is working to enhance the internal architecture of its foundation model.

By refining the model's structure, the company aims to enable it to identify deeper and more complex patterns within payments data.

This enhanced capability could allow the system to detect sophisticated fraud schemes that evolve over time.

The model's ability to recognise emerging patterns could help Mastercard stay ahead of increasingly complex criminal activities.

The technology could also be used to predict market trends and personalise services with greater precision.

These capabilities would enable Mastercard to offer more proactive and relevant services to both merchants and cardholders.

Additionally, Mastercard is training teams across the organisation to access and build applications on top of the foundation model.

This democratisation of AI capabilities could accelerate innovation by enabling product teams, data scientists and business analysts to develop new use cases without requiring extensive machine learning expertise.

For the fintech sector, Mastercard's development roadmap suggests that transaction foundation models could become essential infrastructure for next-generation payment services.