Building Secure AI Infrastructure: Vultr Advances NVIDIA OpenShell and Confidential Computing

As AI agents become more autonomous and frontier models grow in scale, organizations face a shared challenge: how to provide AI workloads with the computing resources they need while protecting sensitive data, proprietary models, and critical systems.

Addressing this challenge requires security at every layer of the infrastructure stack. Vultr is advancing this approach through its work with NVIDIA on agent safety and confidential computing, helping establish the foundation for secure, scalable AI workloads.

Security built into the infrastructure

AI agents increasingly run multi-step workflows, use credentials, call external tools, and interact with internal data with limited human oversight. Traditional safeguards alone cannot reliably constrain what an agent can access or do.

NVIDIA OpenShell, an open-source secure runtime, helps address this challenge by governing and monitoring agents when they operate. The NVIDIA OpenShell Reference Architecture builds on this foundation by adding model compute, governed agent execution, and independent security enforcement through NVIDIA Sentry, a unified security software framework built on NVIDIA DOCA and running on NVIDIA BlueField-4 DPUs.

The architecture brings together three key components:

• NVIDIA Vera Rubin NVL72 runs AI models across training, inference, and agentic workloads.
• NVIDIA Vera CPUs handle agent orchestration, tool use, code execution, and OpenShell secure runtime, which governs agent execution and resource access.
• NVIDIA BlueField-4 DPUs run NVIDIA Sentry, which monitors agent behavior and enforces security policies from a hardware-isolated, host-independent trust domain.

Together, OpenShell and NVIDIA Sentry provide complementary safeguards: runtime governance defines what an agent is permitted to do, while independent hardware-based enforcement helps ensure those policies remain in place.

Protecting sensitive AI workloads with confidential computing

Agent safety is one part of the broader infrastructure security challenge. Organizations also need to protect proprietary models and sensitive data while workloads run on infrastructure they do not own or operate.

NVIDIA Confidential Computing addresses this need through hardware-rooted isolation and cryptographic attestation, helping protect data and models during execution. Its expanded capabilities support three deployment models:

• Confidential Bare Metal: Dedicated infrastructure for sensitive AI workloads.
• Confidential Containers: Kubernetes environments with confidential computing protections.
• Confidential Virtual Machines: Virtualized environments for confidential workloads.

These options give organizations flexibility to align security requirements with their infrastructure and operational needs while accessing accelerated computing resources.

Confidential Computing capabilities are expanding across NVIDIA Grace Blackwell and NVIDIA Rubin systems, including NVIDIA Grace Blackwell NVL72, NVIDIA HGX Rubin NVL8, and NVIDIA DGX Rubin NVL8. NVIDIA DSX OS will support all three confidential computing modes, providing a software foundation for cloud providers building AI factories that support confidential workloads.

Advancing secure AI infrastructure with Vultr

Vultr is building on its NVIDIA Vera Rubin investment with the OpenShell Reference Architecture and supports NVIDIA Confidential Computing deployments. Together, these efforts advance an infrastructure approach that addresses two critical requirements for enterprise AI: controlling what agents can do and protecting sensitive workloads during execution.

For organizations moving AI agents and frontier models into production, these capabilities can help establish clearer security boundaries, improve visibility into agent activity, and protect sensitive workloads across deployment environments.

Open development is also central to this work. OpenShell gives the industry an opportunity to inspect and contribute to the security foundations of agent execution. Through the Open Secure AI Alliance, Vultr encourages infrastructure providers and enterprises to collaborate on advancing secure AI adoption.

As AI systems take on more complex tasks, security must be an integral part of the infrastructure that powers them. By advancing agent safety and confidential computing with NVIDIA, Vultr is helping lay the groundwork for organizations to deploy AI with greater control, isolation, and confidence.

Learn more about NVIDIA Confidential Computing and explore the open-source OpenShell runtime through NVIDIA’s developer resources and GitHub.