AI agents carry memory and storage like people, so their data needs guarding while in use

As AI agents become more prevalent in production on Kubernetes, cloud engineering teams in the financial sector face a new security challenge. While encryption can protect data at rest and in transit, it does not address the pressing issue of securing data while it is being actively used. This is particularly critical for payment networks, where each transaction carries sensitive customer information and regulatory risks.

Krishna Mayuram, senior principal engineer and vice president of cloud engineering at Mastercard, highlighted the importance of "confidential computing," which involves running applications in highly secure virtualized environments or confidential containers to protect data as it is being processed. Mastercard operates its platforms on Red Hat's OpenShift, a Kubernetes platform that has been extended to support AI agents.

Red Hat has also introduced confidential containers to OpenShift on bare metal servers. Like microservices, AI agents require the same elastic infrastructure that Kubernetes provides, including horizontal and vertical scaling. However, scaling the platform increases the risk, especially considering the multi-jurisdictional nature of Mastercard's operations.

To address this, Mastercard has introduced trust services for agentic commerce, applying the same security principles from development to deployment. The company emphasizes the need for consistent guardrails across all areas of its business, whether using AI agents or traditional applications.