NVIDIA 发布 OpenShell 参考架构，保障大规模 AI Agent 工作负载安全
原标题：How NVIDIA OpenShell Reference Architecture Secures and Governs AI Agent Workloads at Scale
AI 摘要
NVIDIA 发布开源运行时 OpenShell 及其参考架构，用于在基础设施层为 AI Agent 工作负载提供安全与治理，将 Agent 视为不可信对象并限制其可访问的系统与数据。OpenShell 包含 Gateway、Supervisor、Sandbox 和 Policy prover 四个组件，运行在 NVIDIA Vera CPU 上。NVIDIA 称 Vera 在关键 Agent 工作负载上性能可达 x86 的 2 倍，CoreWeave 实测沙箱启动速度提升 3 倍以上、单机架支持超过 11000 个并发 Agent 环境。
正文节选
Autonomous agents can work for hours on complex tasks, accessing files, tools, the internet, and spawning other agents to collaborate with. Executing complex tasks over long time horizons leads to compounding risk: the agents may download code, build new tools, request additional privileges, and touch more of the underlying infrastructure. The agents’ action paths are determined as they work and reason through the steps needed to complete their goals. As they interact with the tools and systems