SANTA CLARA, CALIFORNIA / RankWire.AI / – NVIDIA has launched the Open Agent Safety Platform, a security framework designed for autonomous artificial intelligence agents. The system combines software controls with standalone hardware monitoring for AI workloads. It supports development, testing, and deployment, including enterprise computing and robotics. NVIDIA explained that the platform can impose restrictions beyond the AI model itself. Companies can choose individual components based on their security and computing needs.
OpenShell supplies the platform’s open source runtime layer, managing how AI agents access digital resources. Administrators can define rules for file, network, tool, process, and credential access. The software logs agent activities and enforces permissions during task execution. OpenShell creates a security boundary around agents without depending on instructions within the models. NVIDIA designed the software to work with both open-source and proprietary AI models.
NVIDIA Sentry provides an additional monitoring layer through the company’s BlueField-4 data processing units. The technology operates outside the main software environment of the agent and continuously observes activity. Sentry can isolate an agent within milliseconds if its actions exceed security limits. It also enforces access policies independently of the agent’s compliance. NVIDIA’s DOCA software framework supports functions like identity checks, telemetry, and inspection for security purposes.
Extra controls improve agent security
The design separates AI decision-making from security enforcement systems. OpenShell places agents inside controlled runtime environments with isolation and policies. Sentry monitors these environments with dedicated hardware that the agent doesn’t control. The platform was developed after security researchers identified incidents where agents went beyond intended controls. External enforcement offers operators direct oversight of what agents can access and perform.
More than 100 organizations are involved with technologies related to the Open Agent Safety Platform, according to NVIDIA. Participants include Anthropic, Microsoft, Cisco, CrowdStrike, Dell Technologies, HPE, and Hugging Face. The group also features JPMorganChase, Palantir, Palo Alto Networks, Salesforce, SAP, and ServiceNow. Their work covers cybersecurity, cloud infrastructure, enterprise software, and finance. The platform supports autonomous systems that interact with physical environments and need extra controls beyond software alone.
Open source approach expands hardware compatibility
OpenShell operates on NVIDIA Vera CPUs, created for demanding AI workloads. Developers can extend the open source runtime to systems based on third-party architectures. NVIDIA specifically identified Arm and Intel platforms as potential expansion options. Sentry relies on BlueField-4 hardware for its independent monitoring role. This setup separates runtime controls from the hardware layer that observes and restricts agent activity across various computing environments.
Jensen Huang, NVIDIA founder and CEO, described AI safety and security as full-stack engineering challenges when introducing the platform. NVIDIA positioned the Open Agent Safety Platform as a reference system for securing agents across software, hardware, and infrastructure layers. OpenShell provides the software boundary, while Sentry offers hardware-based oversight. Together, these components enable operators to set permissions, record activities, and enforce restrictions as autonomous AI agents transition from development to operational environments.