Breaking

Security Analysis Reveals Governance Gaps in MCP Server Marketplaces Expose AI Infrastructure to Domain Takeover Risks

October 6, 2026 at 07:14 EDT

Security
Policy & Regulation

A security audit of 15,465 published Model Context Protocol (MCP) servers across three major registries has exposed systemic governance failures that leave AI agents vulnerable to infrastructure hijacking and supply-chain attacks. Researchers at OX Security analyzed the landscape of publicly available MCP servers and found that most marketplaces conduct no vetting of hosted tools, creating a fragmented ecosystem where critical AI infrastructure sits exposed to several classes of risk.

AI INFRASTRUCTURE RISK

MCP Marketplaces: Zero Vetting, Real Takeover Risk

15,465 published MCP servers lack centralized security review, leaving expired domains and geographic residency blind spots.

2.3%

Hostnames no longer resolve

Expired domains available for $4–$12

15.6%

Servers outside US (China, Russia)

0.45%

Proxied through home networks

Domain Takeover: The Cascade Risk

When a domain expires: Any AI agent or deployment that references that hostname may continue to trust it. A new owner can capture traffic, tokens, and requests intended for the original server.

No marketplace vetting: MCP registries do not verify server authenticity, uptime, or ownership, leaving the ecosystem open to supply-chain compromise.

📊 Context: 15,465 MCP servers across 3 public registries analyzed by security researchers. No protocol-level controls for data residency, domain ownership validation, or marketplace gatekeeping.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.