Fortanix extends Confidential AI to Kubernetes via CNCF Containers

Fortanix Inc. extended its Confidential AI platform to support the Cloud Native Computing Foundation’s Confidential Containers project today. This update enables enterprises to execute artificial intelligence workloads on Kubernetes within hardware-based trusted execution environments, addressing security gaps during model inference.

The release targets the exposure that occurs when model weights and input data must decrypt in processor memory. Privileged administrators and cloud operators with infrastructure access could potentially reach these assets at that point. The new integration aims to shield data from such privileged access vectors during active processing.

Technical Implementation and Attestation

Fortanix is developing open-source plugins for the Confidential Containers project, which reached incubating status at the CNCF in July. These plugins verify container authenticity before granting access to encryption keys. Verification runs through the Fortanix Confidential Computing Manager, which attests both CPU and GPU components together.

This joint attestation ensures that inference jobs spanning both processors remain fully covered. The company’s Data Security Manager releases keys only if the attestation result satisfies specific customer security policies. Keys remain within the protected environment and do not exit it, maintaining data confidentiality throughout the workflow.

Survey Data and Strategic Context

Fortanix published survey figures last week regarding how AI leaders view these risks. Atomik Research polled 300 leaders at large U.S. companies on behalf of Fortanix. Ninety-three percent stated that traditional encryption does not protect data while AI systems are actively processing it, highlighting a perceived gap in current defenses.

Security concerns had led 41% of respondents to block deployments completely. Co-founder and CEO Anand Kashyap stated that moving AI into containers does not change the threat landscape, as models still handle sensitive data. With the new plugins, deploying a container into a confidential environment reportedly requires one click on supported hardware and platforms.

What should you do?

Organizations deploying AI on Kubernetes should evaluate whether their current encryption strategies cover data in memory during inference. Review CNCF Confidential Containers documentation to understand hardware attestation requirements. Monitor vendor updates for plugin compatibility with existing infrastructure. Assess if privileged access controls are sufficient for current AI workloads.

What to do and how to stay safe: Fortanix

• Audit current AI inference workflows to identify if data is decrypted in processor memory accessible to privileged operators.
• Evaluate hardware support for trusted execution environments required by Confidential Containers on your Kubernetes clusters.
• Review security policies to determine if traditional encryption meets compliance requirements for sensitive data processing.
• Monitor CNCF incubating projects for stability updates before integrating plugins into production AI workloads.

Step-by-step guide: Cloud Misconfiguration Response: Containment, Recovery and Prevention Steps

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What specific project does Fortanix support with this update?

Fortanix supports the Cloud Native Computing Foundation’s Confidential Containers project to secure Kubernetes workloads.

Why is data vulnerable during AI inference?

Model weights and input data must decrypt in processor memory, potentially exposing them to privileged administrators and cloud operators.

Does traditional encryption protect data during processing?

According to a survey by Atomik Research, 93% of AI leaders at large U.S. companies said traditional encryption does not protect data during processing.