UAE, 29 September 2026: NVIDIA has launched the NVIDIA Open Agent Safety Platform, an open software platform and reference system design aimed at strengthening the security and governance of AI agents across software, compute infrastructure and robotics systems.
The platform is designed to give organizations greater control over long-running autonomous agents, particularly as recent security incidents have highlighted the risk of agents bypassing application-level controls while carrying out assigned tasks.
NVIDIA said the initiative brings together industry, researchers and public-sector organizations to share best practices, align on evaluation methods and support international cooperation around AI agent safety. Jensen Huang, founder and CEO of NVIDIA, said the potential of AI will depend heavily on solving safety and security challenges through full-stack engineering.
Full-Stack Governance for AI Agents
The NVIDIA Open Agent Safety Platform provides governance and control across the software running AI agents, the hardware and compute systems powering them, and the robotics platforms that may execute tasks in the physical world. Organizations can deploy different parts of the platform depending on their security and operational requirements.
A key component is NVIDIA OpenShell, secure runtime software that creates enforceable boundaries for autonomous agents operating on CPUs. OpenShell is designed to control how agents execute tasks across open and closed models, giving enterprises a layer of protection outside the model itself.
OpenShell is optimized for NVIDIA Vera, the company’s CPU designed for agentic AI, while also being extendable to third-party compute platforms, including Arm and Intel.
NVIDIA Sentry Adds Hardware-Level Monitoring
The reference design also includes NVIDIA Sentry, an out-of-band monitoring system running on NVIDIA BlueField-4 DPUs.
Sentry continuously monitors agent activity and can isolate and stop an agent in milliseconds if it attempts to operate outside its defined software boundaries. The system combines threat detection, hardware-based governance, security policy enforcement and data-access protection within an isolated trust domain.
Built on NVIDIA DOCA software, Sentry can inspect agent requests and responses, verify agent identity, provide telemetry and enforce zero-trust access policies across data, tools, APIs and services.
Broad Industry Support
NVIDIA said the platform is already being used or integrated by a wide range of technology, enterprise and public-sector organizations.
Anthropic and NVIDIA have collaborated on additional security layers for Claude Managed Agents, while SpaceXAI is using the platform with Cursor coding agents and Grok models. Scale AI is also working with NVIDIA to incorporate the technology into its agentic infrastructure.
Salesforce has integrated OpenShell with Slack, allowing teams to review agent activity and approve or reject requests for additional permissions. SAP is also embedding OpenShell into Joule Studio runtime and contributing engineering work aimed at advancing interoperability standards.
More than 100 organizations are working with NVIDIA Open Agent Safety Platform technologies, including Accenture, Cisco, CrowdStrike, IBM, Microsoft, ServiceNow, Siemens, Palantir and Palo Alto Networks. Robotics companies such as Figure, Gecko Robotics and Skild AI are also using OpenShell to add safety controls to autonomous systems operating in the physical world.
Financial institutions including Citi and JPMorganChase are collaborating with NVIDIA on open-source agent safety technologies, while energy and infrastructure companies including Hitachi Energy, Schneider Electric, Siemens Energy and Worley are also involved.
Availability
NVIDIA said the Open Agent Safety Platform software, including OpenShell and related skills, is available through NVIDIA developer resources and GitHub.
The initiative also supports the broader work of the Open Secure AI Alliance, which was initiated by NVIDIA alongside more than 120 organizations and is governed by the Linux Foundation, with a focus on open research, tools and collaboration around AI safety and security.
