Mastercard Executive Discusses Protecting AI Agent Data in Use

Mastercard senior principal engineer and vice president of cloud engineering Krishna Mayuram discussed protecting data while AI agents process it at KubeCon + CloudNativeCon North America in Salt Lake City, held Nov. 9-12.

The discussion focused on data “in use” — information software is actively processing. Encryption can protect information while it is stored or moving between systems, but it does not by itself protect data as an application uses it. Mayuram discussed confidential computing as one approach to that challenge; the interview included no announcement of a new Mastercard deployment or product.

Confidential computing uses hardware isolation to protect data during computation. With Confidential Containers on hardware that supports a trusted execution environment, each Kubernetes pod can run in a separate, hardware-isolated virtual machine. Cryptographic verification of runtime integrity can be required before the workload receives secrets.

Red Hat made confidential AI on bare-metal GPUs generally available Aug. 4. It also designated its Red Hat build of Agent Sandbox a Technology Preview. Those announcements do not establish that Mastercard uses either product.

Mastercard separately announced Agent Pay trust and intelligence services Sept. 30. Its first service was a score designed to estimate whether a transaction was initiated by an AI agent, and it was rolling out for testing in the United States. That announcement did not describe confidential computing.