{"version":1,"type":"story","url":"https://digestai.news/story/nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a","json":"https://digestai.news/story/nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a.json","markdown":"https://digestai.news/story/nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a.md","slug":"nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a","headline":"NVIDIA launches OpenShell reference architecture for safe AI agents","summary":"NVIDIA announced the OpenShell Reference Architecture at GTC Berlin on October 21, 2026. The platform is designed to build, contain, observe, and enforce safe AI agents on enterprise infrastructure and edge systems. It combines NVIDIA Vera Rubin infrastructure for model inference, Vera CPUs for agent harnesses, and BlueField-4 DPUs running the Sentry security framework. This setup allows for continuous out-of-band monitoring and in-silicon policy enforcement, which can quarantine agents that attempt to move outside their defined boundaries.\n\nThe architecture supports leading enterprise operating systems, including those from Canonical, Red Hat, and Microsoft. NVIDIA claims Vera Rubin NVL72 systems deliver 30x higher throughput per megawatt for agentic workloads, while Vera CPUs offer 2x the performance of x86 processors. Sentry, powered by DOCA software, boosts agent capacity by up to 20%.\n\nMajor partners including Anthropic, AWS, Cisco, CrowdStrike, Google Cloud, IBM, Meta, Microsoft, OpenAI, and SAP are working with NVIDIA to enhance agent security. ServiceNow announced the general availability of ServiceNow Arc, a personal AI agent governed by NVIDIA OpenShell. The architecture is now available for deployment across cloud and on-premises environments.","keyPoints":["NVIDIA OpenShell combines Vera Rubin, Vera CPUs, and BlueField-4 DPUs for secure agent execution.","Vera Rubin NVL72 systems deliver 30x higher throughput per megawatt for agentic workloads.","ServiceNow Arc is generally available, using OpenShell for runtime governance and security enforcement."],"whyItMatters":"This architecture provides a standardized hardware and software foundation for safely deploying AI agents at scale. It addresses critical security concerns by enabling real-time monitoring and containment, which is essential for enterprise adoption of autonomous AI systems.","category":{"slug":"hardware","name":"Hardware & Compute","url":"https://digestai.news/category/hardware"},"entities":{"companies":["NVIDIA","Anthropic","AWS","Cisco","CrowdStrike","Google Cloud"],"models":[],"people":["Jensen Huang"]},"firstPublishedAt":"2026-10-21T02:47:00Z","updatedAt":"2026-10-21T10:13:11Z","sourceCount":4,"hasPrimarySource":false,"sources":[{"outlet":"manilatimes.net","title":"NVIDIA Launches OpenShell Reference Architecture for Safe AI Agents","url":"https://manilatimes.net/2026/10/21/tmt-newswire/globenewswire/nvidia-launches-openshell-reference-architecture-for-safe-ai-agents/2450556","publishedAt":"2026-10-21T02:51:00Z","type":"press","primary":false,"lead":true},{"outlet":"Unite.AI","title":"NVIDIA Debuts OpenShell Reference Architecture for Agent Governance","url":"https://unite.ai/nvidia-debuts-openshell-reference-architecture-for-agent-governance","publishedAt":"2026-10-21T10:13:11Z","type":"press","primary":false,"lead":false},{"outlet":"cryptobriefing.com","title":"Nvidia and Anthropic partner to enhance AI security with OpenShell integration","url":"https://cryptobriefing.com/nvidia-and-anthropic-partner-to-enhance-ai-security-with-openshell-integration","publishedAt":"2026-10-21T03:05:00Z","type":"press","primary":false,"lead":false},{"outlet":"247wallst.com","title":"NVIDIA OpenShell Makes AI Agent Safety a Silicon Problem for Enterprises","url":"https://247wallst.com/cards/nvidia-launches-openshell-reference-architecture-for-safe-ai-agents-nvda-press-release-01m5eyj6f7nyr77sdk6m446dg7","publishedAt":"2026-10-21T02:47:00Z","type":"press","primary":false,"lead":false}],"sourceNotes":null,"discussions":[],"thread":{"title":"Nvidia Eyes Reflection AI Amid Open Model Launch","url":"https://digestai.news/thread/reflection-may-release-open-weight-model-to-rival-chinese-ai","storyCount":6},"cite":{"text":"Digest AI, \"NVIDIA launches OpenShell reference architecture for safe AI agents\", 21 October 2026, https://digestai.news/story/nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a","publisher":"Digest AI","title":"NVIDIA launches OpenShell reference architecture for safe AI agents","datePublished":"2026-10-21T02:47:00Z","url":"https://digestai.news/story/nvidia-launches-openshell-reference-architecture-for-safe-enterprise-a"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}