📄 中文摘要
自主代理标志着人工智能的一个新转折点。系统不再仅限于生成响应或推理任务，而是能够采取行动。代理可以读取文件、使用工具、编写和运行代码，并在企业系统中执行工作流程，同时不断扩展自身的能力。随着代理的持续改进，应用层风险呈指数级增长，因此需要在设计阶段就考虑安全性，以确保这些自主系统的可靠性和安全性。
NVIDIA OpenShell 如何使自主 AI 代理在设计上安全
出处: How Autonomous AI Agents Become Secure by Design With NVIDIA OpenShell
发布: 2026年3月23日
📄 English Summary
How Autonomous AI Agents Become Secure by Design With NVIDIA OpenShell
Autonomous agents represent a new inflection point in AI, moving beyond merely generating responses or reasoning through tasks to taking actionable steps. These agents can read files, utilize tools, write and execute code, and manage workflows across enterprise systems while continuously enhancing their capabilities. As agents improve over time, the application-layer risks grow exponentially, necessitating a focus on security by design to ensure the reliability and safety of these autonomous systems.