SANTA CLARA, CALIFORNIA / RankWire.AI / – NVIDIA has launched the Open Agent Safety Platform, a comprehensive security framework designed for autonomous artificial intelligence agents. This system integrates software safeguards with dedicated hardware monitoring across various AI applications, spanning development, testing, and operational deployment, including enterprise computing and robotics sectors. NVIDIA emphasized that the platform can impose restrictions beyond the AI models themselves, allowing organizations to select specific components aligned with their security and computational needs.
OpenShell serves as the open-source runtime layer that manages how AI agents access digital resources. It enables administrators to create rules governing file, network, tool, process, and credential access. While tasks are active, the software logs agent activities and enforces permissions, establishing a security boundary around agents without relying on embedded instructions within the models. NVIDIA designed OpenShell to support both open and proprietary AI models effectively.
The NVIDIA Sentry component introduces an additional security layer via the company’s BlueField-4 data processing units. This technology operates independently of an agent’s primary software environment, continuously monitoring activity. Sentry can isolate an agent within milliseconds if its actions breach predefined security parameters and can enforce access policies autonomously, without relying on the agent for compliance. NVIDIA’s DOCA software framework underpins these security functions, supporting identity verification, telemetry, inspection, and more.
Decoupled controls enhance agent protection
The platform’s architecture distinctly separates AI decision-making from the systems that implement security policies. OpenShell encapsulates agents within controlled runtime environments with built-in isolation and enforcement, while Sentry monitors these environments via dedicated hardware managed separately from the agents. This design stems from security research revealing cases where agents bypassed intended application controls. NVIDIA asserts that external enforcement provides operators with direct oversight of agent access and execution capabilities.
More than 100 organizations are engaged with technologies related to the Open Agent Safety Platform, including Anthropic, Microsoft, Cisco, CrowdStrike, Dell Technologies, HPE, and Hugging Face. The list also features JPMorganChase, Palantir, Palo Alto Networks, Salesforce, SAP, and ServiceNow. Their efforts span cybersecurity, cloud infrastructure, enterprise software, and financial services. The platform also underpins autonomous systems that interact with physical environments, requiring controls beyond software alone.
Open source approach expands hardware compatibility
OpenShell is compatible with NVIDIA Vera CPUs, designed for high-demand AI workloads involving agents. Developers can extend its open-source runtime to third-party hardware architectures, with support explicitly identified for Arm and Intel platforms. Sentry relies on BlueField-4 hardware for its independent monitoring functions, enabling a separation of runtime controls from the hardware layer responsible for observing and restricting agent activity across compatible systems.
Jensen Huang, NVIDIA’s founder and CEO, described AI safety and security as full-stack engineering challenges at the platform’s announcement. The company positioned the Open Agent Safety Platform as a benchmark reference for securing agents across software, hardware, and infrastructure layers. OpenShell offers the software boundary, while Sentry provides hardware-based independent oversight. These combined elements empower operators to set permissions, track activity, and enforce restrictions as autonomous AI agents transition from development to operational deployment.