Critical Flowise Flaw Enables Remote Command Execution via MCP Adapters

OX Security disclosed a critical, architecture-level vulnerability in Anthropic’s Model Context Protocol (MCP) that allows unauthenticated arbitrary command execution (RCE) across official MCP SDKs (Python, Rust, Java, TypeScript) and many downstream projects. The flaw yields access to internal databases, user data, chat histories, and API keys, has resulted in multiple CVEs across AI products, and poses a large supply-chain risk; Anthropic has declined to change the protocol, so mitigations and patches must be applied by downstream organizations.