SANTA CLARA, CALIFORNIA / RankWire.AI / – NVIDIA has rolled out the Open Agent Safety Platform, a comprehensive security framework designed for autonomous artificial intelligence agents. This system integrates both software-based controls and standalone hardware monitoring to oversee AI workloads. Its scope spans the entire lifecycle from development and testing to deployment, including systems utilized in enterprise computing and robotics. NVIDIA emphasized that the platform can impose restrictions beyond what is embedded within the AI models themselves. Organizations are also able to select specific components aligned with their security and computing needs.
OpenShell serves as the platform’s open-source runtime layer, managing how AI agents access digital resources. It allows administrators to define rules governing access to files, networks, tools, processes, and credentials. The software logs agent activities and enforces permissions during task execution. OpenShell establishes a security boundary around agents without depending on embedded instructions within the models. It is compatible with both open and proprietary AI models, NVIDIA explained.
NVIDIA Sentry introduces an additional monitoring layer through the company’s BlueField-4 data processing units. The technology operates independently of an agent’s core software environment, continuously observing activity. When an agent’s actions breach predefined security parameters, Sentry can isolate the agent within milliseconds. It also enforces access policies without relying on the agent itself for compliance. NVIDIA’s DOCA software framework supports identity verification, telemetry, inspection, and other security functionalities.
Segregated controls bolster agent security
This architecture isolates AI decision-making from security enforcement systems. OpenShell places agents inside protected runtime environments that feature isolation and policy management. Sentry then monitors these environments via dedicated hardware that the agent does not manage. The platform was developed in response to security researchers’ reports of agents exceeding their intended application controls. NVIDIA stated that external enforcement mechanisms allow operators to have direct oversight over what agents can access and perform.
According to NVIDIA, over 100 organizations are working with technologies related to the Open Agent Safety Platform. Participants include Anthropic, Microsoft, Cisco, CrowdStrike, Dell Technologies, HPE, and Hugging Face. The collaboration also involves JPMorganChase, Palantir, Palo Alto Networks, Salesforce, SAP, and ServiceNow. These groups operate across cybersecurity, cloud infrastructure, enterprise software, and financial services sectors. The platform also supports autonomous systems that interact with physical environments and require controls beyond software applications.
Open source architecture enhances hardware compatibility
OpenShell is compatible with NVIDIA Vera CPUs, designed specifically for demanding agentic AI workloads. Developers have the ability to extend the open source runtime to systems based on third-party architectures, with NVIDIA identifying Arm and Intel platforms as supported expansion routes. Sentry relies on BlueField-4 hardware for its independent monitoring functions. This setup separates runtime controls from the hardware layer responsible for observing and restricting agent activity across compatible computing environments.
Jensen Huang, NVIDIA’s founder and CEO, described AI safety and security as full-stack engineering challenges during the platform’s announcement. The company positions the Open Agent Safety Platform as a reference model for securing agents across software, hardware, and infrastructure layers. OpenShell provides the software boundary, while Sentry offers independent hardware oversight. Collectively, these components empower operators to establish permissions, log activities, and enforce restrictions as autonomous AI agents transition from development to operational environments.