Nvidia launches its OpenShell Reference Architecture, a framework to secure and monitor AI agents, partnering with Canonical, AWS, Google, Microsoft, and others
Nvidia had already framed agent deployment as a trust and safety problem in its 2025 inference microservices push. Its September Open Agent Safety Platform reference design paired OpenShell for CPUs with Sentry for Nvidia DPUs to contain agents within sandboxes.
The new architecture extends that work from containment toward a broader operating model for securing, monitoring and governing agents. It also gives practical form to Nvidia's Open Secure AI Alliance, which was formed in July to develop and share AI-safety and cybersecurity tools.
First-order effects
Nvidia gives Canonical, AWS, Google and Microsoft a common reference architecture around which to build or validate AI-agent deployments with security, monitoring and governance controls.
Enterprise teams evaluating Nvidia-based agent systems gain a defined infrastructure-level design for handling agent safety rather than relying solely on application-layer controls.
Second-order effects
Nvidia's cloud and software partners are pushed to show how their services implement the architecture's controls, making governed agent operations a differentiator in enterprise AI deployments.
Security vendors participating around OpenShell gain a clearer integration point with Nvidia infrastructure, while buyers can assess agent controls alongside compute and platform choices.
Third-order effects
If partners adopt the reference design broadly, AI-agent security may consolidate around infrastructure-enforced isolation, identity and monitoring rather than fragmented controls added by individual applications.
The move fits a market in which AI infrastructure suppliers package compute, software and governance as one stack, increasing the strategic value of interoperable security standards.
The trend: AI-agent deployment is moving toward integrated infrastructure stacks where governance and security are designed into the runtime alongside compute.
As AI agents become more autonomous, security needs to be built into the infrastructure powering them. NVIDIA's new OpenShell Reference Architecture helps make this possible, with Arm-based Vera CPUs powering agent execution and BlueField DPUs supporting isolation and policy enfo…
We are proud to support the @NVIDIAAI OpenShell Reference Architecture and bring it to our infrastructure. Building on GMI Cloud's investment in the NVIDIA Vera Rubin platform: model compute, governed agent execution and hardware-level security are all on one cloud. Agents can…
Moving AI agents from pilot to production takes more than speed, it takes control. Together with @NVIDIAAI OpenShell, Palo Alto Networks secures the AI factory with governed identities, isolated runtimes, and real-time SOC signals. See how. https://www.paloaltonetworks.com/ ...
We are glad to announce our plans to work with @nvidia to explore the potential of the OpenShell Reference Architecture to enable new approaches to AI agent safety, security and governance. The collaboration is part of our strategy to scale the use of AI agents across the organiz…
As @JensenHuang said, AI safety is a technological issue. This is the right approach, and one that helps hold companies accountable. If a plane is unsafe, can it operate? NO. So if AI labs can't guarantee their products are safe, they shouldn't be able to operate either.
Agent security belongs in the infrastructure layer. We're proud to partner with @nvidia on the OpenShell Reference Architecture, an open architecture for running AI agents with security enforced at the infrastructure level. We build safety into how we serve models. Base Labs is b…
“You should first start with zero trust” when using an autonomous agent, says Jensen Huang. Nvidia revealed its OpenShell Reference Architecture at its GTC Berlin event. The Nvidia CEO sat down with @dee_bosa. “The way to think about safety is we have to accelerate AI safety,” [v…
AI agents need more than guardrails. CrowdStrike and @NVIDIAAI are extending AI agent security from software to silicon. Announced at #NVIDIAGTC Berlin: Falcon Guardian is being integrated with NVIDIA OpenShell, with plans for in-silicon enforcement through NVIDIA Sentry. 🔗 Learn…