Nvidia推出了OpenShell 0.1.x版本，这是一个专为自主AI代理设计的安全私密运行时工具1。该系统通过内核级隔离和形式化验证技术，在允许代理访问文件、安装包、调用API和使用凭证的同时，防止其获得对数据、密钥和网络的无限制访问1。OpenShell采用沙箱隔离和策略检查机制来控制代理行为1。
该版本包含稳定发布节奏、新隔离原语、扩展表面和新API1。在数据隐私方面，OpenShell不收集沙箱名称、主机名、文件路径、提示、凭证、提供商或模型名称等敏感信息1。目前OpenShell支持Linux、macOS on Apple Silicon和Windows with WSL 2（实验阶段）等系统1，并采用Apache License 2.0开源许可证1。
Nvidia has unveiled OpenShell 0.1.x, a secure and privacy-focused runtime environment designed specifically for autonomous AI agents 1. The platform employs kernel-level isolation and formal verification techniques to enable agents to access files, packages, APIs, and credentials while preventing unrestricted access to sensitive data, cryptographic keys, and network resources 1.
The runtime leverages sandbox isolation and policy enforcement mechanisms to control agent behavior 1. OpenShell 0.1.x introduces a stable release cadence, new isolation primitives, an expanded surface area, and new APIs 1. The system operates on Linux, macOS on Apple Silicon, and Windows with WSL 2 in an experimental phase 1. Notably, the platform does not collect sandbox names, hostnames, file paths, prompts, credentials, provider information, or model names, prioritizing user privacy 1. Nvidia has released OpenShell under the Apache License 2.0 open-source license 1.
评论
还没有评论，欢迎留下第一条。