NVIDIA Launches Open Agent Safety Platform for Secure AI Development and Deployment

NVIDIA has announced the launch of its Open Agent Safety Platform, a comprehensive solution designed to strengthen AI security from agent testing to deployment. The platform consists of NVIDIA’s OpenShell open-source software and the NVIDIA Sentry reference system design, which enables full-stack governance and control across software and hardware systems that run agents. This move marks an important step towards ensuring the safe development and deployment of artificial intelligence (AI) technologies in various industries.

The need for robust AI safety measures has become increasingly pressing due to recent security incidents involving long-running agents. These incidents have highlighted the importance of equipping organizations with open, customizable tools that enforce more control over agent behavior. The pattern observed across these incidents is consistent – the agent circumvented security controls at the application layer to complete its assigned task.

NVIDIA’s Open Agent Safety Platform addresses this critical need by providing a secure runtime boundary for controlling how autonomous AI agents execute tasks across various systems and models. This platform enables organizations to deploy elements according to their unique requirements, ensuring that safety measures are tailored to specific needs. The solution includes NVIDIA OpenShell, which sets boundaries for agents running on CPUs, and Sentry, an out-of-band watchdog that continuously monitors agent behavior.

Sentry is a crucial component of the platform, as it provides in-silicon security enforcement by quarantining and stopping AI agents that attempt to move outside their software boundary. This feature ensures that safety policies are enforced independently in silicon, combining threat detection, hardware-based agent governance, and data access protection from an isolated trust domain. Sentry’s capabilities are built on NVIDIA DOCA software, which provides programmable capabilities for inspecting agent requests and responses, verifying agent identity, and enforcing granular zero-trust access policies.

The Open Agent Safety Platform has garnered significant attention from industry leaders across the AI ecosystem. Over 100 organizations have joined forces with NVIDIA to strengthen AI safety through shared open-source technologies. These collaborations include Anthropic, Cisco, CrowdStrike, Dell Technologies, Figure, HPE, Hugging Face, JPMorganChase, Microsoft, Palantir, Palo Alto Networks, Perplexity, Red Hat, Salesforce, SAP, Scale AI, ServiceNow, and SpaceXAI.

Anthropic has collaborated with NVIDIA to bring additional layers of security and control to the agent stack through Claude Managed Agents. This solution establishes a security boundary by running the agent loop in a separate server from the sandboxes where their work executes. Integrations with OpenShell and BlueField enable enterprises to enforce strict control over agent access.

SpaceXAI is utilizing NVIDIA’s Open Agent Safety Platform for its Cursor coding agents and Grok models, emphasizing the importance of enforcing safety measures outside the model by additional controls that agents cannot bypass. Scale AI is working closely with NVIDIA to incorporate platform technologies into its agentic infrastructure layer, ensuring reliable operation in mission-critical applications.

Salesforce has integrated OpenShell with Slack, enabling teams to manage agent activity directly from the platform – viewing agent activity and audit events, approving or rejecting requests for additional permissions. SAP is embedding OpenShell within Joule Studio runtime as part of its Business AI Platform, pairing business oversight with runtime security. The company is also contributing engineering work to OpenShell and collaborating with NVIDIA on advancing interoperability standards through the Open Secure AI Alliance.

The launch of the Open Agent Safety Platform marks a significant milestone in addressing the pressing need for robust AI safety measures. As Jensen Huang, founder and CEO of NVIDIA, noted, ‘AI’s extraordinary potential for society will only be realized if we solve AI safety.’ The platform brings together industry leaders to share best practices, align on evaluation methods, and foster international cooperation – ultimately raising the bar for global AI safety.

NVIDIA Open Agent Safety Platform software is available through the company’s developer resources page and GitHub. This move supports the mission of the Open Secure AI Alliance, which aims to strengthen AI agent security through open research, skills, and tools. The alliance has been initiated by NVIDIA alongside over 120 leading organizations and governed by the Linux Foundation.

The availability of this platform is expected to have a significant impact on various industries, including finance, energy, and infrastructure software. Citi and JPMorganChase are among financial services leaders collaborating with NVIDIA on shared open-source agent safety technologies. Energy leaders such as Hitachi Energy, EPRI, NextEra Energy, Quanta Services, SPP, Schneider Electric, Siemens Energy, and Worley are also working closely with NVIDIA to integrate platform technologies into their operations.

Infrastructure software leaders Canonical, SUSE, and Red Hat are integrating the Open Agent Safety Platform into widely used operating systems. Red Hat is running OpenShell and DOCA on its AI Factory with NVIDIA – a co-engineered enterprise-grade AI solution for building, deploying, and managing AI at scale across hybrid cloud environments.

NVIDIA’s partners, including Baseten, Cisco, CoreWeave, Dell Technologies, GMI Cloud, HPE, HP Inc., Irregular, Lenovo, Microsoft, Nebius, Oracle Cloud Infrastructure, Supermicro, and Together AI, are offering AI infrastructure solutions that use and support NVIDIA Open Agent Safety Platform technologies to help customers run AI agents more securely.