Anthropic's Model Context Protocol includes a critical remote code execution vulnerability — newly discovered exploit puts 200,000 AI servers at risk

Researchers found a critical remote code execution flaw in Anthropic’s Model Context Protocol (MCP) related to STDIO execution in the reference SDKs.

Unsanitized user input can reach command execution, enabling prompt-injection and command-injection exploits across multiple platforms.

The issue has already led to several high-severity CVEs and could affect hundreds of thousands of AI servers and downstream tools.

OX Security says the risk is broad because many MCP-based apps, registries, and AI coding tools may inherit the vulnerability.

Anthropic has not yet released a protocol-level fix, leaving a significant supply-chain security concern.