Mastercard's AI-Driven Evolution: Redefining Fraud Prevention and Trust

In a groundbreaking shift, Mastercard is redefining its approach to fraud prevention in a world where artificial intelligence (AI) agents are becoming legitimate participants in the buying process. Traditionally, fraud systems were designed to view bots as potential thieves, but as these intelligent agents gain acceptance in the marketplace, the rules of engagement are evolving dramatically. At the heart of this transformation is Greg Ulrich, Mastercard's Chief AI and Data Officer, who recently shared insights during a session at VB Transform 2026 in Menlo Park, California. His remarks underscore a fundamental change in the financial landscape: as AI agents move from the periphery to the center of commerce, the systems designed to protect transactions must also adapt.

With a staggering 175 billion transactions processed in the last year alone, Mastercard's fraud detection system operates under a time constraint of less than 100 milliseconds to assess and score the likelihood of fraud. This rapid analysis is crucial for maintaining trust in a digital economy where consumers expect seamless, secure transactions. Ulrich emphasized the importance of this trust, as it forms the backbone of commerce, enabling merchants and consumers alike to engage in transactions with confidence. The challenge now lies in recalibrating the risk frameworks to accommodate AI agents as legitimate buyers rather than potential threats.

The Shift to AI-Enabled Transactions

Mastercard's evolution from viewing bots as adversaries to enabling them as buyers reflects a broader trend in commerce where AI plays an integral role. Ulrich stated that the company has spent decades building risk rules designed to thwart fraudulent bots. However, with the advent of generative AI and enhanced data processing capabilities, the focus is shifting towards empowering legitimate AI agents to transact while simultaneously safeguarding against fraud.

Enhanced Fraud Detection Capabilities

The integration of generative AI into Mastercard's fraud detection system has led to significant improvements in identifying potential fraud. According to Ulrich, the new technology allows Mastercard to process more data and context, resulting in the detection of 300% to 400% more fraudulent transactions at high-risk levels without adding unnecessary friction for consumers. The company's Safety Net system has already prevented over 70 billion fraudulent transactions, underscoring the effectiveness of this new approach.

Building Trust Through Robust Frameworks

As AI agents become more prevalent in transactions, the concept of trust is paramount. Ulrich outlined a multi-layered framework that Mastercard has developed to ensure the integrity of AI-enabled commerce. This framework consists of five key layers:

• Identity Verification: Understanding both the consumer and the AI agent involved in the transaction.
• Verifiable Intent: Maintaining a tamper-proof record of the original transaction instructions.
• Control Mechanisms: Defining merchant access, transaction limits, and constraints for AI agents.
• Execution Framework: Utilizing Mastercard's Agent Pay system to manage authentication and acceptance.
• Intelligence Layer: Incorporating risk rules and insight tokens to monitor and personalize transactions.

This comprehensive approach not only addresses the complexities of AI-driven transactions but also lays the groundwork for greater trust between consumers and businesses. By ensuring that every transaction can be traced back to clear intent and identity, Mastercard is paving the way for a safer digital commerce environment.

The Future of Agentic Commerce

The implications of this shift extend beyond consumer purchases to encompass business-to-business (B2B) transactions, where the potential for automation is even greater. Ulrich provided an illustrative example of a manufacturing scenario where an AI agent manages inventory levels, tracks stock, and makes procurement decisions autonomously. This level of automation requires the same five layers of trust and verification developed for consumer transactions, but on a larger scale.

Addressing Multi-Party Trust

To facilitate B2B transactions involving multiple parties—such as suppliers, banks, and procurement agents—Mastercard recognizes the need for standardized protocols that ensure a seamless flow of information and trust. The complexity of these transactions necessitates clear standards for identity verification and intent validation, as well as mechanisms for communication between various agents.

As businesses increasingly adopt AI technologies for procurement and supply chain management, Mastercard's framework could serve as a model for establishing trust in agentic commerce. This evolution signifies a critical shift in how businesses manage transactions, making it essential for stakeholders to understand both the opportunities and the inherent risks involved.

Lessons Learned and Future Directions

Mastercard's journey over the past 14 months has provided valuable insights into the integration of AI within its operational framework. Ulrich shared that the key lessons learned include the necessity of embedding security measures at the outset rather than as an afterthought, the importance of scaling operations effectively, and the critical nature of observability and accountability within AI systems.

Rethinking Architecture

In retrospect, Ulrich noted that had Mastercard been starting from scratch, the company would approach the design of AI systems fundamentally differently to account for the evolving landscape of digital commerce. The rapid pace of change in AI technology and the emergence of new vulnerabilities highlight the need for a proactive rather than reactive strategy in security and trust-building.

Key Takeaways

• Mastercard is adapting its fraud prevention strategies to accommodate AI agents as legitimate buyers.
• The company’s Safety Net system has prevented over 70 billion fraudulent transactions.
• Five layers of trust have been established to secure AI-driven transactions.
• Agentic commerce has the potential to transform B2B procurement through automation.
• Embedding security at the design phase is crucial for successful implementation.

Frequently Asked Questions

How is Mastercard changing its fraud detection approach?

Mastercard is evolving its fraud detection approach by shifting its focus from viewing bots as potential threats to enabling them as legitimate buyers. This transformation involves recalibrating risk frameworks and incorporating advanced technologies, such as generative AI, to enhance the detection of fraudulent transactions while maintaining user experience.

What are the five layers of trust in Mastercard's transaction framework?

The five layers of trust include identity verification, verifiable intent, control mechanisms, execution frameworks, and an intelligence layer. These components work together to ensure that transactions involving AI agents are secure, transparent, and reliable, thereby fostering greater trust between consumers and businesses.

What lessons has Mastercard learned from implementing AI technologies?

Mastercard has learned the importance of embedding security measures during the design phase of AI systems, operating at scale, and ensuring accountability in AI decision-making processes. These lessons are shaping the company's future strategies and approaches to digital commerce.

How does agentic commerce impact B2B transactions?

Agentic commerce has the potential to revolutionize B2B transactions by allowing AI agents to automate procurement processes, manage inventory, and facilitate transactions autonomously. However, this shift requires a robust framework for trust and verification to ensure all parties involved can confidently engage in automated transactions.