Model Context Protocol faces security scrutiny over server vulnerabilities

The Model Context Protocol (MCP), designed to standardize how AI models interact with data, is facing significant security scrutiny. At the MCP Dev Summit in Seoul, experts addressed a growing landscape of vulnerabilities. Research indicates that over 21,000 internet-facing MCP server instances are currently exposed, with nearly 92% of audited production servers lacking basic OAuth authentication.

A central debate exists regarding the protocol’s STDIO transport model. While OX Security identified systemic architectural vulnerabilities affecting millions of downstream package downloads, Anthropic has maintained that the current behavior is ‘by design,’ asserting that the execution model serves as a ‘secure default’ and that developers are responsible for input sanitization.