NVIDIA releases open-source agent safety platform with chip-level continuous monitoring and isolation
Original title:NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring
AI summary
NVIDIA launched the NVIDIA Open Agent Safety Platform, including the open-source runtime OpenShell, the hardware-layer Sentry and DOCA technology. The platform aims to prevent AI agent behavior drift and privilege escalation through kernel-level isolation, zero-trust environments and out-of-band monitoring. OpenShell converts operational instructions into verifiable policies, while BlueField DPUs provide real-time policy enforcement and identity governance on the model path, ensuring independent system protection even when the host is untrusted.
Key Takeaways
- NVIDIA launched the NVIDIA Open Agent Safety Platform,including the open-source runtime…
- The platform aims to prevent AI agent behavior drift and privilege escalation through kernel-level…
- OpenShell converts operational instructions into verifiable policies,while BlueField DPUs provide…
Quick Facts
- Published
- Updated
- Category
- Products & Apps
- AI score
- 82
针对前沿实验室报告的智能体逃逸风险,NVIDIA 提供了从软件到硬件的完整安全栈方案。其“带外监控”和“芯片级隔离”理念为构建可信 AI 基础设施提供了重要参考,适合关注 Agent 安全落地的开发者与企业。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: NVIDIA releases open-source agent safety platform with chip-level continuous monitoring and isolation. September 28, 2026. https://new.nat6.net/en/news/s5on5dsnixevj7j9lfjemhz3o/