Over 200,000 AI Servers Exposed to Attack Risks: MCP Revealed to Have Severe Design Flaws
The cybersecurity company OX Security reported that the MCP protocol of Anthropic has a design flaw, and its SDK's STDIO interface could lead to remote code execution, affecting over 200,000 AI servers. As an open-source standard, MCP is widely used in AI application development, and this vulnerability has attracted significant attention from the industry.