📄 中文摘要
大语言模型（LLMs）正在迅速成为与数据交互的新接口层。用户通过自然语言提问，期望获得实时且准确的答案。然而，这一转变带来了一个关键挑战：将LLM连接到数据库或API时，实际上将其转变为动态数据访问层。如果没有适当的控制，这一层可能会成为安全和治理风险。为了解决这一问题，提出了在LLM驱动系统中实施真实数据治理的方法，重点在于可以立即应用的实用模式。
构建安全的对话式人工智能：大语言模型驱动接口的数据治理模式
📄 English Summary
Building Secure Conversational AI: Data Governance Patterns for LLM-Powered Interfaces
Large Language Models (LLMs) are rapidly emerging as a new interface layer for data interaction, allowing users to ask questions in natural language and expect real-time, accurate responses. However, this shift presents a critical challenge: connecting an LLM to databases or APIs effectively turns it into a dynamic data access layer. Without proper controls, this layer can pose significant security and governance risks. The article outlines methods for implementing effective data governance in LLM-powered systems, focusing on practical patterns that can be applied immediately.