Research · curated 21 Jul 2026

Rethinking MCP Security: A Large-Scale Study of Runtime MCP Servers and Security Scanner Reliability

Coverage timeline

14 Jul 2026arxiv.orgprimary 20 Jul 2026securitypointbreak.com

Why it matters

MCP security scanners that defenders rely on to vet AI-agent tool servers are wrong more than half the time, meaning organizations may trust unreliable risk verdicts about the plumbing connecting LLM agents to external tools.

A large-scale study, "Rethinking MCP Security" (arXiv:2607.11086), presents MCPZoo, the largest collection of MCP servers for dynamic analysis (64,611 unique servers, 37,288+ runnable), and measures the reliability of MCP security scanners. While existing scanners flag 96.89% of servers as risky, the authors find fewer than 50% of sampled alerts are true positives and that scanner outputs are inconsistent across tools.