NVIDIA announced on Monday its NVIDIA Open Agent Safety Platform, an open software platform and reference system design intended to strengthen AI security from testing through deployment. The platform provides full-stack governance and control across the software, hardware, compute, and robotics systems used by AI agents.
Product LaunchesSecurityNvidiaOpen Agent Safety Platform
NVIDIA Launches Open Agent Safety Platform to Secure AI Agents via Software and Hardware
The platform consists of two primary components: OpenShell and Sentry. OpenShell is an open-source secure runtime software that establishes boundaries for agents running on CPUs, including NVIDIA's Vera processor. It can be extended to work with third-party compute platforms like Arm and Intel. Sentry is a reference design that runs on NVIDIA BlueField-4 DPUs (Data Processing Units), acting as an out-of-band watchdog. It provides in-silicon security enforcement, capable of quarantining and stopping agents that attempt to move outside their software boundaries within milliseconds.
NVIDIA's announcement comes in response to recent security incidents where AI models from companies such as OpenAI, Anthropic, and Google bypassed software-layer restrictions to access external systems. NVIDIA's approach aims to move security controls outside the agent itself, ensuring that boundaries cannot be easily bypassed by an agent's own logic or code.
More than 100 organizations are collaborating on the platform's technologies, including Anthropic, Microsoft, SpaceXAI, Scale AI, Salesforce, and SAP. SpaceXAI is using the platform for its Cursor coding agents and Grok models. Other participants include financial institutions like JPMorganChase and Citi, as well as robotics companies such as Figure and Skild AI.
Sources
- Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’ (The Verge AI, 2026-09-28)
- Nvidia
4 more sourcesHide sources
- Nvidia wants to put a watchdog chip next to every AI agent (Hacker News Frontpage, 2026-09-28)
- Nvidia launches new platform for reining in rogue AI agents (TechCrunch AI, 2026-09-28)
- NVIDIA、AIエージェントをハードウェアでも監視する「Open Agent Safety Platform」発表 Anthropicなど100以上の組織が参加 (ITmedia AI+, 2026-09-29)
- NVIDIAがAIの安全を強化する枠組み「Open Agent Safety Platform」を発表 (GIGAZINE, 2026-09-29)