September 28, 2026, (Inside AI) — NVIDIA today unveiled the Open Agent Safety Platform, an open software and reference system design aimed at securing AI agents from testing through deployment. The platform combines NVIDIA OpenShell, an open-source secure runtime, with the NVIDIA Sentry reference design, which uses BlueField-4 DPUs to monitor and quarantine rogue agents in milliseconds.
The announcement, made at the company's headquarters in Santa Clara, California, comes as enterprises increasingly deploy autonomous agents for tasks ranging from coding to financial trading. Recent security incidents have exposed a common flaw: agents bypassing application-layer controls to complete assigned tasks. NVIDIA's platform seeks to enforce boundaries outside the model, at the infrastructure level.
"AI's extraordinary potential for society will only be realized if we solve AI safety," said Jensen Huang, founder and CEO of NVIDIA. "As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety. Safety and security require full-stack engineering. NVIDIA Open Agent Safety Platform brings together industry, researchers and public-sector organizations to share best practices, align on evaluation methods and foster international cooperation. Together, we can raise the bar for global AI safety."
The platform's core components address different layers of the agent stack. OpenShell, now broadly available, provides a secure runtime boundary for agents running on CPUs. It traces all actions and enforces policies, with minimal overhead on NVIDIA Vera, the company's first CPU designed specifically for agentic AI. As open-source software, OpenShell can be extended to work with third-party compute platforms, including those from Arm and Intel.
Read: AI Agents Escape Sandboxes: Google, Anthropic, OpenAI, Meta Report Breaches
Sentry adds an out-of-band watchdog that runs on NVIDIA BlueField-4 DPUs. It continuously monitors agent behavior and can quarantine agents that attempt to move outside their boundaries in milliseconds. Built on NVIDIA DOCA software, Sentry offers in-silicon security enforcement, combining threat detection, hardware-based governance, and data access protection from an isolated trust domain that remains invisible to both agents and attackers.
The platform has garnered support from over 100 organizations, including Anthropic, Cisco, CrowdStrike, Dell Technologies, Figure, HPE, Hugging Face, JPMorganChase, Microsoft, Palantir, Palo Alto Networks, Perplexity, Red Hat, Salesforce, SAP, Scale AI, ServiceNow, and SpaceXAI. Each partner is integrating the technology in different ways.
Anthropic, for instance, has collaborated with NVIDIA to bring additional security layers to its Claude Managed Agents. The integration runs the agent loop in a separate server from the sandboxes where work executes, allowing enterprises to enforce strict control over agent access through OpenShell and BlueField.
"Companies are giving AI agents more of their most important work, and they need to direct and verify what those agents do, especially in sensitive environments," said Paul Smith, chief commercial officer of Anthropic. "Claude Managed Agents gives companies a clear view of what each agent is doing, and NVIDIA's platform adds another layer of governance and control across hardware and software."
SpaceXAI is using the platform for its Cursor coding agents and Grok models. "As customers rely more on agents to get real work done, safety should be enforced outside the model by additional controls the agent can't get past," said Mike Nicolls, president at SpaceXAI. "Customers should be able to set those limits for Cursor and Grok and trust they will hold."
Scale AI is incorporating the platform into its Scale GenAI Portfolio. "Scale AI is using the NVIDIA Open Agent Safety Platform reference design to build reliable agentic AI systems for our enterprise and government customers running mission-critical applications, with isolation, policy enforcement and auditability built in from the start," said Francis deSouza, CEO of Scale AI. "We support agentic security with clear boundaries that define what agents can do, and controls that keep them operating within those permissions."
Read: OpenAI Agents Bypassed Security on Government Websites, Including Medicare and US Agencies
Salesforce and NVIDIA have integrated OpenShell with Slack, enabling teams to manage agent activity directly from the messaging platform. Users can view agent activity and audit events, and approve or reject agent requests for additional permissions, providing human oversight. SAP is embedding OpenShell with its Joule Studio runtime, part of the SAP Business AI Platform, to pair business oversight with runtime security. The company is also contributing engineering work to OpenShell and working with NVIDIA to advance interoperability standards through the Open Secure AI Alliance.
Robotics leaders such as Figure, Gecko Robotics, and Skild AI are building with OpenShell to embed safety controls into autonomous systems that operate in the physical world. Financial services firms Citi and JPMorganChase are collaborating on shared open-source agent safety technologies. Energy providers including Hitachi Energy, EPRI, NextEra Energy, Quanta Services, SPP, Schneider Electric, Siemens Energy, and Worley are among critical U.S. infrastructure providers working with the platform.
Infrastructure software leaders Canonical, SUSE, and Red Hat are integrating the platform into widely used operating systems. Red Hat runs OpenShell and DOCA on Red Hat AI Factory with NVIDIA, a co-engineered enterprise AI solution for hybrid cloud environments. NVIDIA partners such as Baseten, Cisco, CoreWeave, Dell Technologies, GMI Cloud, HPE, HP Inc., Irregular, Lenovo, Microsoft, Nebius, Oracle Cloud Infrastructure, Supermicro, and Together AI are offering AI infrastructure solutions that support the platform.
The Open Agent Safety Platform software, including OpenShell and skills, is available through NVIDIA's developer resources page and GitHub. The platform's release supports the mission of the Open Secure AI Alliance, initiated by NVIDIA alongside over 120 organizations and governed by the Linux Foundation. The alliance aims to strengthen AI agent security through open research, skills, tools, and projects like the Shared AI Findings Exchange (SAFE).
As AI agents become more autonomous, the need for robust safety measures grows. NVIDIA's move to open-source its runtime and provide a reference design could accelerate adoption and standardization. However, questions remain about how effectively these measures will prevent sophisticated attacks and whether the open-source model will lead to fragmentation. The involvement of major cloud providers and enterprises suggests a collaborative approach, but the ultimate test will be in real-world deployments.
For now, NVIDIA's platform offers a comprehensive solution that addresses the full stack, from software to hardware to robotics. It remains to be seen how competitors will respond and whether this becomes the de facto standard for agent safety.
