Independent Scan of 5,618 MCP Servers: SSRF, Path Traversal, and 2010-Era Web Vulnerabilities Dominating AI Infrastructure

DEV Community·high signal

A recently published independent security analysis of 5,618 MCP servers found that critical vulnerabilities predominantly map to well-understood 2010-era web security classes — SSRF, path traversal, injection, and unsafe deserialization — not novel AI-specific attack vectors. The findings challenge the assumption that AI infrastructure requires exotic new tooling, suggesting existing AppSec scanners adapted for MCP tool schemas would catch most issues without AI-specialized security products. This dataset is distinct from the BlueRock MCP Trust Registry study (8,000 servers, same 36.7% SSRF finding) and provides complementary technical depth on the same systemic failure pattern.