AI代理的隐患：模型上下文协议（MCP）中的11个关键安全风险

The Model Context Protocol (MCP), released by Anthropic in November 2024, has quickly become the standard for connecting Large Language Models (LLMs) to external tools and data sources. With widespread adoption across major platforms such as Claude Desktop, OpenAI Agent SDK, Microsoft Copilot Studio, Amazon Bedrock Agents, Cursor, and Visual Studio Code, MCP now handles millions of requests daily. However, this rapid growth has introduced a critical attack surface that many organizations fail to recognize. Our analysis reveals 11 distinct vulnerabilities within MCP that pose significant threats to the security of systems and data utilizing this protocol.