SANTA CLARA, CALIFORNIA / RankWire.AI / – NVIDIA has unveiled the Open Agent Safety Platform, a comprehensive security solution designed for autonomous artificial intelligence agents. This system integrates software controls with dedicated hardware monitoring to oversee AI workloads. It is applicable throughout the development, testing, and deployment stages, including systems used in enterprise computing and robotics. NVIDIA emphasized that the platform can impose restrictions beyond the AI model itself, allowing organizations to select specific components based on their computing and security needs.
The platform’s open source runtime layer, called OpenShell, manages how AI agents access digital resources. It enables administrators to establish rules governing access to files, networks, tools, processes, and credentials. OpenShell logs agent activity and enforces permissions during task execution. Its design creates a security boundary around agents without depending on embedded instructions within the models. NVIDIA ensured compatibility with both open and proprietary AI models by developing this software to function across different types of systems.
Monitoring is further enhanced by NVIDIA Sentry, which provides an additional layer through the company’s BlueField-4 data processing units. The technology operates independently of an agent’s main software environment, continuously observing activity. Sentry can isolate an agent within milliseconds if its actions violate security parameters. Furthermore, it enforces access policies without relying on the agent for compliance. To support identity verification, telemetry, inspection, and security functions, NVIDIA utilizes its DOCA software framework.
Separate controls bolster agent security
The architecture distinctly separates AI decision-making from the systems tasked with enforcing security policies. OpenShell creates controlled runtime environments where agents operate with isolation and policy enforcement. Meanwhile, Sentry monitors these environments using dedicated hardware that the agent itself does not control. The platform was developed in response to security researchers who documented cases of agents exceeding their intended application controls. NVIDIA highlighted that external enforcement provides operators with direct oversight over what agents can access and execute.
More than 100 organizations are engaged with technologies related to the Open Agent Safety Platform, according to NVIDIA. This group includes Anthropic, Microsoft, Cisco, CrowdStrike, Dell Technologies, HPE, and Hugging Face. The consortium also features JPMorganChase, Palantir, Palo Alto Networks, Salesforce, SAP, and ServiceNow. Their collective efforts span cybersecurity, cloud infrastructure, enterprise software, and financial services. The platform additionally supports autonomous systems interacting with physical environments that demand controls beyond software applications.
Open source approach expands hardware compatibility
OpenShell is compatible with NVIDIA Vera CPUs, which are designed for intensive agentic AI workloads. Developers also have the option to extend the open source runtime to third-party architectures. NVIDIA has specifically identified Arm and Intel platforms as supported options for expansion. Sentry relies on BlueField-4 hardware for its independent monitoring role, effectively separating runtime controls from the hardware layer responsible for observing and restricting agent activity across compatible environments. NVIDIA founder and CEO Jensen Huang described AI safety and security as full-stack engineering challenges during the platform’s announcement. The Open Agent Safety Platform is positioned as a reference system to secure agents across software, hardware, and infrastructure layers. OpenShell defines the software boundary, while Sentry provides independent hardware oversight. Together, these components enable operators to set permissions, log activity, and enforce restrictions as autonomous AI agents transition from development to operational environments.