Fortanix extends Confidential AI to Kubernetes workloads through Confidential Containers

Data security firm Fortanix Inc. has expanded its Fortanix Confidential AI platform to incorporate the Cloud Native Computing Foundation's Confidential Containers project. This allows enterprises to securely run AI workloads on Kubernetes within hardware-based trusted execution environments. The focus is on securing data during inference, when model weights and associated data must be decrypted in processor memory, potentially exposing it to privileged administrators and cloud operators.

According to a recent survey by Atomik Research, 93% of AI leaders believe traditional encryption does not protect data during processing, with 41% blocking deployments altogether due to security concerns. Fortanix's solution leverages Confidential Containers, which reached incubating status at the CNCF in July. The company's plugins verify container authenticity before granting access to encryption keys, ensuring the entire inference job, spanning both CPU and GPU, remains protected.

Fortanix's Data Security Manager only releases keys if they meet the customer's security policy. The update is designed for regulated industries and can be deployed on-premises or fully air-gapped. Major companies such as AMD, Intel, and CrowdStrike are supporting the effort. Nvidia is collaborating with Fortanix on an on-premises confidential AI platform.

Fortanix Confidential AI for Confidential Containers is now available, and the company plans to showcase it at Nvidia GTC Berlin on October 20-22.