Model Context Protocol’s STDIO flaw exposes 200K AI servers to remote code execution
A critical design default in Anthropic’s Model Context Protocol allows unchecked command execution across 200,000 MCP-linked servers. Researchers found no input sanitization, enabling attackers to bypass security controls and execute arbitrary commands without detection.
