Rafay and Fortanix bring confidential AI to neoclouds and enterprises

Rafay Systems and Fortanix have announced a joint solution for delivering confidential AI as a service on neocloud, sovereign AI cloud and enterprise infrastructure. The offering uses hardware based Trusted Execution Environments, cryptographic attestation and secure key release to protect data and AI models while they are being processed.
- Rafay and Fortanix announced a joint confidential AI solution.
- CoCo runs Kubernetes pods in lightweight confidential virtual machines.
- Fortanix releases encryption keys only to attested environments.
- Users launch confidential workloads through the Rafay Developer Hub.
- Support for CoCo based deployments is available now.
AI workloads run inside attested environments
The Rafay Platform now orchestrates workloads running on CNCF Confidential Containers, or CoCo, with Fortanix Confidential AI. The integration extends the companies' existing support for confidential virtual machines.
CoCo applies confidential computing to the Kubernetes pod. Built on Kata Containers, it runs each pod in its own lightweight confidential virtual machine, with memory encrypted by the CPU. On supported hardware, GPUs are included in the trust boundary.
Fortanix Confidential Computing Manager performs composite attestation across the environment. It verifies the integrity of CPU and GPU trusted execution environments, firmware and software components before authorizing encryption key release. Fortanix Data Security Manager, backed by a FIPS 140-3 Level 3 HSM certification, releases keys only to environments that have passed attestation. If verification fails, the keys remain protected.
Rafay adds multi tenant delivery and self service
Rafay turns the underlying security controls into a service that operators can offer to customers. Rafay blueprints install Fortanix CoCo add-ons and keep them synchronized across confidential clusters. Operators can onboard GPU infrastructure once, then provide each customer with an isolated tenant.
Administrators can package confidential workloads as SKUs in Rafay SKU Studio. End users launch them on demand through the Rafay Developer Hub without filing a support ticket. Policy, access controls and quotas apply across tenants and clusters.
According to Rafay chief product officer Mohan Atreya, the arrangement allows a neocloud or sovereign cloud to offer confidential AI as a SKU, while model providers can deploy proprietary models on infrastructure they do not operate. Fortanix chief product officer Anuj Jaiswal said the system is designed so model owners, enterprises and operators can each retain control of what they own.
Use cases span regulated industries and AI cloud providers
The companies point to financial services, healthcare and the public sector as key use cases. These sectors may process customer records, patient information and financial transactions, while model providers also need to protect proprietary model weights.
Neoclouds and GPU cloud providers can offer confidential AI as a premium service backed by attestation evidence. Sovereign AI clouds can keep keys in an HSM under the operator's control and verify what is running before data is decrypted. Regulated enterprises can process sensitive data while keeping it protected from the infrastructure operator during processing.
Fortanix and Atomik Research found that 93% of AI leaders understand traditional encryption does not protect data while AI is processing it, while only half can fully verify that protection is in place during inference. Rafay support for CoCo based deployments with Fortanix Confidential AI is available now, alongside the existing confidential VM integration.


