Executive Summary
OX Security analyzed 15,465 public MCP servers and describes widespread security problems in the ecosystem around the protocol.
What Happened
In 2024, MCP (Model Context Protocol) set out to become the USB-C of AI: one standard for connecting models, agents, and IDEs to tools and data. The protocol delivered. Thousands of developers built servers, and enterprises plugged them into agent workflows. The ecosystem around it fell short. Earlier this year, our team at OX Security, traced critical vulnerabilities in Anthropic's MCP
Affected Products
- Model Context Protocol servers
Vulnerability / Attack Technique
Insecure public MCP servers
Technical Mechanism
Research survey of server security issues.
AI Security Category
LLM03:2025 Supply Chain
Part of the MCP Security topic.
Part of the AI Supply Chain and AI SBOM topic.
Relevant Frameworks
RealGround Analysis
The MCP ecosystem is large and uneven. RealGround analysis: allowlist MCP servers rather than trusting the ecosystem.
Recommended Controls
- Allowlist vetted MCP servers
Related RealGround Service
AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/10/welcome-to-jungle-what-we-found-inside.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.