Mastercard Engineer Discusses Securing Data for AI Agents

At KubeCon + CloudNativeCon North America in Salt Lake City, a Mastercard engineer addressed protecting data while AI agents process it.

Updated on Nov. 12, 2026 in Artificial Intelligence

Mastercard Engineer Discusses Securing Data for AI Agents

Mastercard senior principal engineer Krishna Mayuram discussed protecting data while AI agents process it at KubeCon + CloudNativeCon North America, held November 9-12 in Salt Lake City. Encryption protects information at rest and in transit but does not by itself protect data while an application uses it.

Why it matters

Encryption alone leaves data exposed during computation, a gap that becomes more significant as AI agents handle sensitive information. Approaches such as confidential computing aim to close that gap by isolating data in hardware during processing.

Confidential Containers can run each Kubernetes pod in a separate hardware-isolated virtual machine on compatible hardware that supports a trusted execution environment. Cryptographic verification of runtime integrity can be required before a workload receives secrets, though the source does not quantify overhead.

The players

Mastercard

Mastercard is a global payments company that operates one of the largest card networks in the world and has been developing services for AI-driven transactions.

Red Hat

Red Hat is an enterprise software company owned by IBM that develops open-source platforms including Kubernetes-based cloud infrastructure products.

The details

Krishna Mayuram explained that confidential computing uses hardware isolation to protect data during computation rather than only at rest or in transit. Mastercard's first Agent Pay service is a score estimating whether a transaction was initiated by an AI agent, now rolling out for testing in the United States.

Timeline

  1. August 4: Red Hat made confidential AI on bare-metal GPUs generally available.

  2. September 30: Mastercard announced Agent Pay trust and intelligence services.

  3. November 9-12: KubeCon + CloudNativeCon North America was held in Salt Lake City.

The Tech Race

The push to isolate AI workloads in hardware follows the broader cloud-native movement that moved containers from simple shared hosts toward stronger isolation. Mastercard's Agent Pay score and Red Hat's confidential AI releases show competitors racing to secure agent-driven transactions.

Users of AI-powered payment tools may see transactions flagged or scored as agent-initiated once Mastercard's Agent Pay score reaches broader testing. Confidential computing could eventually protect personal data handled by AI services, though availability depends on compatible hardware.

The takeaway

Protecting data is no longer only about encrypting it at rest and in transit; the moment an AI agent uses it, hardware-level isolation becomes the relevant safeguard to watch.

Further reading

For more on the technology behind this, explore Artificial Intelligence coverage on Wisevoter.

Source note: This article includes information reported by TokenPost.