Anthropic MCP Design Vulnerability Enables RCE, Threatening AI Supply Chain

Cybersecurity researchers from OX Security disclosed a critical, architecture-level weakness in the Model Context Protocol (MCP) STDIO configuration that permits arbitrary command execution on systems running vulnerable MCP implementations. The flaw, present in Anthropic's reference SDKs across Python, TypeScript, Java, and Rust and inherited by many downstream projects (including LiteLLM, LangChain, LangFlow, Flowise and others), has produced multiple CVEs and affects thousands of publicly reachable servers and widely downloaded packages; some vendors have patched, but the protocol design remains unchanged, leaving a broad supply-chain risk.