
Key points
- Encryption alone cannot secure data while AI agents process it in Kubernetes environments.
- Mastercard uses Red Hat OpenShift to run AI workloads with confidential container capabilities.
- Guardrails must span development and deployment to meet diverse international regulatory requirements.
SiliconANGLE reported that Krishna Mayuram, senior principal engineer and vice-president of cloud engineering at Mastercard Inc, discussed the security challenges of running AI agents on Kubernetes during an exclusive broadcast at KubeCon + CloudNativeCon NA.reported He stated that protecting data while it is being processed, rather than just at rest or in motion, has become central to running agents at scale for financial companies.He stated that protecting data while it is being processed, rather than just stored or moving, is now essential for large-scale AI agent operations.He stated that protecting data in use within virtualised environments is now critical for scaling AI agents on Kubernetes infrastructure.
Evidence: reported
AI agents carry memory and storage like people, so their data needs guarding while in useaifleet-siliconangle
The evidence
Mayuram told theCUBE Research’s John Furrier and Rebecca Knight that payment networks face heightened pressure because every transaction carries customer data and regulatory exposure. He explained that container orchestration has become the default foundation for AI workloads, making the data those agents use the main target for security threats. SiliconANGLE noted that cloud engineering teams at financial companies face a problem that traditional encryption methods alone cannot solve.
The decision
The engineer described confidential computing as the ability to protect data in use by launching applications in highly secure, virtualised environments or confidential containers. He contrasted this with previous security models that focused on data at rest and data in motion, asking what happens to data while it is actively being used. This shift requires engineering teams to adopt secure, virtualised environments specifically designed for processing sensitive information within containerised infrastructure.
The takeaway
Mastercard runs its platforms on Red Hat Inc’s OpenShift, which Red Hat is extending to support AI agents and has brought confidential containers to on bare metal. Mayuram explained that agents act much like microservices and require the same elastic infrastructure capabilities, including horizontal and vertical scaling. He noted that scale increases risk, particularly for a company operating under many countries’ laws, necessitating guardrails at every step from development to deployment.
Building on this
Developers can build AI agents that leverage Kubernetes’ elastic capabilities for scaling, treating them similarly to microservices in terms of infrastructure requirements. They must integrate confidential computing techniques to protect data in use within virtualised secure environments, as standard encryption is insufficient for processing stages. Implementation requires applying consistent guardrails across distributed systems to comply with varying international laws, ensuring security from development through to deployment.
Why it matters
- Enterprise IT buyers must evaluate if their Kubernetes platforms support confidential containers for data-in-use protection.
- Security strategies need to expand beyond data-at-rest and data-in-motion encryption to cover active processing phases.
- Compliance requirements for distributed financial services demand guardrails that adapt to multiple international legal frameworks.
- Scaling AI agents introduces new security risks that traditional microservice approaches may not adequately address without confidential computing.
Topics: Kubernetes, confidential computing, AI agents, data security, Mastercard
Updated
Then / now / next
Security focused on protecting data at rest and in motion using encryption.
Confidential computing protects data in use within virtualised secure environments for AI agents.
Engineering teams must implement guardrails across distributed systems to meet diverse regulatory requirements.
Story timeline
- Then Security focused on protecting data at rest and in motion using encryption.
- 11 Nov 2026 AI agents carry memory and storage like people, so their data needs guarding while in use
- Now Confidential computing protects data in use within virtualised secure environments for AI agents.
- Next Engineering teams must implement guardrails across distributed systems to meet diverse regulatory requirements.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
Confidential computing
Protecting data in use by launching applications in highly secure, virtualised environments or confidential containers.
Kubernetes
A container orchestration platform that provides elastic infrastructure capabilities for scaling applications and AI agents.
Confidential containers
Secure, virtualised environments within container orchestration that protect data while it is being processed.
Guardrails
Security measures and controls applied at every step from development to deployment to ensure compliance.
Questions readers ask
What is the primary security challenge for AI agents on Kubernetes?
Protecting data while it is being processed, which traditional encryption methods do not adequately cover.
Which platform does Mastercard use for its AI workloads?
Mastercard runs its platforms on Red Hat Inc’s OpenShift, which supports AI agents and confidential containers.
How do AI agents differ from traditional applications in infrastructure needs?
Agents act like microservices and require the same elastic infrastructure capabilities, such as horizontal and vertical scaling.
Connected stories
Reading receipt
- Encryption alone cannot secure data while AI agents process it in Kubernetes environments.
- Mastercard uses Red Hat OpenShift to run AI workloads with confidential container capabilities.
- Guardrails must span development and deployment to meet diverse international regulatory requirements.
Still open: How do confidential containers impact the performance or latency of AI agent workloads?
Still open: What specific technical steps are required to implement these guardrails in existing Kubernetes deployments?