Skip to content
Vulnerabilities in Model Context Protocol Expose Home Servers to Attacks

Vulnerabilities in Model Context Protocol Expose Home Servers to Attacks

First seen 19 Oct 2026, 04:20 UTC • •
API

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 19, 2026 at 05:49 UTC
  • •0.45% of scanned MCP servers are unsecured, posing a significant risk.
  • •CVE-2025-6514 is a critical vulnerability with a CVSS score of 9.6, likely to be exploited.
  • •Anthropic's design choice leaves developers responsible for sanitizing commands in MCP SDKs.

A report by OX Security reveals that 0.45% of 15,465 scanned Model Context Protocol (MCP) servers are running on unsecured personal machines, creating significant risks for users. The vulnerabilities stem from a command injection flaw in the MCP SDKs, confirmed by Anthropic as a design choice, which leaves developers responsible for sanitization. Two critical vulnerabilities, CVE-2025-6514 and CVE-2026-30624, have been identified, with CVE-2025-6514 having a CVSS score of 9.6 and a high likelihood of exploitation. The architecture of MCP allows for potential exploitation by AI worms, which could self-replicate across IoT devices. Users of MCP should be aware that not all implementations carry the same risk, with Google MCP offering a more secure alternative. The report highlights the urgent need for developers to address these vulnerabilities to prevent exploitation.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated just now How this analysis works

Timeline

2025-07-09
CVE-2025-6514 published
A critical remote code execution vulnerability in the mcp-remote npm package was disclosed, impacting many MCP clients.
Tech.Yahoo
2026-04-15
CVE-2026-30624 published
A high-severity remote code execution vulnerability in Agent Zero was disclosed, allowing arbitrary command execution.
Tech.Yahoo

More articles in this cluster (3)

Following this threat?

Track CVE-2025-6514 in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Extracted Entities (18)

1 / 4

Common questions

How many MCP servers are affected?
Approximately 200,000 MCP server instances are estimated to be vulnerable globally.
What should developers do to mitigate risks?
Developers should implement proper command sanitization to address the command injection vulnerabilities.
Are there safer alternatives to MCP?
Yes, Google MCP uses a more secure architecture that mitigates these risks.