Nvidia has launched the Open Agent Safety Platform, giving organisations security controls across the software, compute and hardware layers that underpin autonomous AI agents.
CEO Jensen Huang said AI safety requires controls beyond the model itself.
“AI’s extraordinary potential for society will only be realised if we solve AI safety,” Huang said in a 28 September statement.
“As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety.”
The platform includes Nvidia OpenShell, an open-source runtime software that establishes an enforceable boundary around autonomous agents running on CPUs designed to control how agents execute tasks across both open and closed models, addressing a security gap where controls can potentially be bypassed.
OpenShell runs on Nvidia Vera CPUs with minimal overhead and can also be extended to third-party compute platforms, including Arm and Intel systems.
The reference architecture also includes Nvidia Sentry, an “out-of-band watchdog” running on BlueField-4 data processing units. Sentry continuously monitors agent behaviour independently of the agent itself and can quarantine and stop an agent attempting to breach its software boundary in milliseconds.
Nvidia said the platform is intended to give organisations customisable controls as agents become more autonomous and operate across longer-running workflows while also providing a common framework for researchers, industry and public-sector organisations to develop and evaluate AI safety practices.
Several other AI firms, including Anthropic and SpaceXAI, are working with Nvidia on the project. In Anthropic’s case, Claude Managed Agents establishes a security boundary by running an agent loop on a separate server.
“Companies are giving AI agents more of their most important work, and they need to direct and verify what those agents do, especially in sensitive environments,” Paul Smith, chief commercial officer of Anthropic, said.
“Claude Managed Agents gives companies a clear view of what each agent is doing, and Nvidia’s platform adds another layer of governance and control across hardware and software.”
TrendAI is another company working closely with Nvidia to support its new platform.
"Agentic AI is the most significant shift in enterprise technology in a generation, and security needs and guardrails are in place to keep the agents aligned with the intentions of the organisation,” Rachel Jin, chief platform and business officer, head of TrendAI, said.
“Security can’t sit in one place. It has to cover the model, the harness, and every tool and data source an agent touches. Nvidia is building the boundary that determines what an agent can do. TrendAI brings the threat intelligence and oversight that tell security teams what that boundary should be, and what is happening inside it. Together, we give organisations a practical path to scale agentic AI securely.”
Accenture, Armadin, Cadence, Cognition, CrowdStrike, Cisco, Dassault Systèmes, Deloitte, EY, Hugging Face, IBM, Irregular, Perplexity, Microsoft, SAP, Scale AI, ServiceNow, Siemens, Synopsys, OpenClaw, Palantir and Palo Alto Networks are among more than 100 organisations working with Nvidia on its new platform.
Want to see more stories from trusted news sources?Make Cyber Daily a preferred news source on Google.
