📄 中文摘要
系统提示是 AI 产品的核心，它们编码了产品逻辑、差异化和指令，构成了 AI 功能背后的“如何”。然而，许多团队在向第三方 API 提供商发送请求时，未对这些系统提示进行加密或编辑，导致系统提示泄露问题。这一问题结构性强、普遍存在，且大多数团队尚未对此进行深入思考。系统提示的内容通常包括公司名称、融资阶段等关键信息，这些信息的泄露可能对企业的竞争力和隐私安全造成严重影响。
系统提示泄露问题：您的 AI 产品架构暴露于您使用的每个提供商
📄 English Summary
The System Prompt Leak Problem: Your AI Product Architecture Is Exposed to Every Provider You Use
System prompts are the crown jewels of AI products, encoding product logic, differentiation, and instructions that define the 'how' behind AI features. Many teams send these prompts unencrypted and unredacted to third-party API providers with every request, leading to the system prompt leak problem. This issue is structural, widespread, and largely overlooked by teams. A typical production system prompt may include critical information such as company name and funding stage, and the leakage of this information can severely impact a company's competitive edge and privacy security.