Anthropic MCP Design Vulnerability Enables RCE, Threatening AI Supply Chain

Anthropic's Model Context Protocol contains a critical design flaw enabling remote code execution through unsafe STDIO transport defaults, affecting over 150 million downloads across multiple languages.

• •Arbitrary command execution vulnerability allows attackers to access sensitive user data, API keys, databases, and chat histories from vulnerable MCP implementations
• •Over 7,000 public servers and 150+ million downloads affected across Python, TypeScript, Java, and Rust language implementations
• •Four attack categories identified: unauthenticated command injection via STDIO, direct STDIO configuration bypass, zero-click prompt injection, and MCP marketplace exploits
• •Anthropic considers the behavior "expected" and has declined to modify the protocol, leaving developers responsible for mitigations