SANTA CLARA, CALIFORNIA / RankWire.AI / – NVIDIA has launched its Open Agent Safety Platform to strengthen security controls for autonomous AI agents. The platform covers agent testing, deployment, software, computing infrastructure and robotics systems. NVIDIA announced the technology on September 28 as enterprises expand their use of agents across complex computing environments. The system combines open source runtime software with independent hardware monitoring. NVIDIA said organizations can deploy individual platform components according to their security and operational requirements.
OpenShell serves as the platform’s secure runtime layer and creates enforceable boundaries around AI agents. The open source software traces agent activity and applies policies while agents run on CPUs. Operators can define access to files, networks, tools, processes and credentials before an agent begins work. OpenShell then enforces those limits during operation. NVIDIA said the software supports open and closed AI models and provides a security boundary outside the model and its agent harness.
The second major component, NVIDIA Sentry, provides independent monitoring through BlueField-4 data processing units. Sentry operates outside the agent’s software environment and continuously watches agent behavior. The system can quarantine an agent within milliseconds when it attempts to move beyond its permitted boundary. It also enforces security policies independently through hardware. NVIDIA built Sentry on its DOCA software, which supports agent identity checks, access controls, telemetry and inspection of agent requests and responses.
Runtime controls extend beyond the model
The platform links application software, secure runtime controls and computing infrastructure into a layered security architecture. OpenShell places agents inside sandboxed environments with kernel-level isolation and policy enforcement. Sentry adds a separate hardware layer that agents cannot directly control. NVIDIA said recent security incidents showed agents bypassing application-level controls while completing assigned tasks. The company designed the system to give organizations controls that remain separate from the AI agent and the model directing its actions.
More than 100 organizations are working with technologies connected to the Open Agent Safety Platform, according to NVIDIA. Anthropic has collaborated with the company on additional controls for managed AI agents and sandbox environments. Salesforce has integrated OpenShell with Slack so teams can review agent activity, audit events and permission requests. Other participants span financial services, enterprise software, cybersecurity, cloud infrastructure, robotics and energy. The platform also supports security controls for autonomous systems that operate in physical environments.
Open source software expands deployment options
OpenShell runs with NVIDIA Vera, the company’s CPU designed for agentic AI workloads. NVIDIA also allows developers to extend the open source software to third-party computing platforms, including systems using Arm and Intel technology. The broader platform can operate across compatible hardware environments, while Sentry uses BlueField-4 for its independent monitoring functions. NVIDIA said Open Agent Safety Platform software, including OpenShell and related skills, is available through its developer resources and the GitHub software development platform.
NVIDIA founder and CEO Jensen Huang said the company views AI safety and security as full-stack engineering challenges. The platform also contributes technology to the Open Secure AI Alliance, an initiative governed by the Linux Foundation. That alliance focuses on open research, tools and security practices for AI agents. NVIDIA said its platform supports governance across software, hardware, computing and robotics layers. The release adds new runtime and hardware controls as organizations deploy autonomous AI agents across a growing range of business systems.