Nvidia partners with AWS and Canonical on OpenShell framework for AI agent governance

Nvidia has introduced the OpenShell Reference Architecture, a framework designed to secure, monitor, and govern artificial intelligence agents within corporate environments. According to a report by The Register, the initiative aims to balance agent utility with strict operational constraints. The launch involves partnerships with Canonical, Amazon Web Services, and BBVA, signaling a coordinated effort to establish standards for enterprise AI deployment.

What is the OpenShell Reference Architecture?

OpenShell is a reference architecture proposed by Nvidia to keep corporate AI agents in line. The framework provides a structure for securing and monitoring these agents, ensuring they operate within defined boundaries. It addresses the operational challenge of allowing AI systems to perform useful work while simultaneously locking down their execution environment to prevent unauthorized actions or data leakage.

Who are the partners involved?

The initiative includes collaborations with several major technology and financial institutions. Canonical, a key player in Linux server infrastructure, is listed as a partner. Amazon Web Services is also involved, bringing cloud infrastructure expertise to the framework. Additionally, BBVA, a global financial group, participates, indicating early interest in applying these governance controls within regulated industries that require strict compliance.

What problem does this framework solve?

Corporate AI agents require freedom to execute tasks but pose significant security risks if left unchecked. Nvidia’s approach seeks to constrain agent execution to mitigate these risks. By providing a reference architecture, the company offers a blueprint for organizations to implement necessary security controls without hindering the functional capabilities of their AI systems, addressing a growing concern in enterprise IT operations.

What does this mean for operations?

Organizations deploying AI agents can use this architecture as a guide for implementation. The partnership model suggests that the framework is intended for broad industry adoption rather than being a proprietary Nvidia solution. Operations teams can expect to see more standardized approaches to agent governance, potentially simplifying the process of integrating AI tools into existing security and compliance workflows.

Questions to ask your suppliers

Determine whether your current AI vendors provide clear boundaries for agent execution. Ask how they monitor agent behavior in real time and what logging capabilities exist for audit trails. Identify who within your organization is responsible for defining acceptable use policies. Ensure that legal and compliance teams review any third-party frameworks before adoption. Check if suppliers offer reference architectures that align with your internal security standards. Verify that data handling procedures remain transparent and controllable throughout the agent lifecycle.

Background: AI agents

AI agents are software systems that perceive their environment, make decisions, and execute actions to achieve specific goals. They move beyond generating text to performing tasks like coding, scheduling, or data analysis. You must treat them as active participants in your network, not passive tools.

Read the full guide: AI Agents Examples: How Autonomous Software Changes Security

What to do and how to stay safe: Nvidia

• Review current AI agent deployments for uncontrolled execution paths.
• Assess whether existing monitoring tools can track agent actions effectively.
• Define clear operational boundaries for AI tasks within your environment.
• Consult with vendors on available governance frameworks and controls.

Step-by-step guide: Browser Security Settings for Small Business Teams

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

Frequently asked questions

What is Nvidia OpenShell?

It is a reference architecture launched by Nvidia to secure, monitor, and govern corporate AI agents, developed in partnership with firms like AWS and Canonical.

Which companies are partnering with Nvidia on this?

Partners include Canonical, Amazon Web Services, and BBVA, according to reports from The Register.

Does OpenShell restrict AI agent functionality?

The framework aims to allow agents to do useful work while locking them down to ensure they remain in line with corporate governance standards.