MT.1120: AI agents should not use MCP server tools without review.
| ID | MT.1120 |
| Severity | Medium |
| Suite | Mito |
| Product | Copilot Studio |
| Feature | Agents |
| Security area | AI and agents |
| Resource type | Agent |
| User impact of fixing | Low |
| Effort to fix | Medium |
| MITRE ATT&CK | T1195, T1199 |
| Reads from | Microsoft Graph |
What this checks and how to fix it
Section titled “What this checks and how to fix it”AI agents should not use MCP server tools without review.
Model Context Protocol (MCP) tools extend agent capabilities by connecting to external servers. These integrations introduce supply chain risks — if an MCP server is compromised, tools-poisoned or untrusted, it could provide malicious instructions, exfiltrate data, or execute unauthorized actions through the agent.
How to fix
Section titled “How to fix”Review all MCP server integrations in the flagged agents. Ensure each MCP server endpoint is owned by your organization or a trusted partner, is hosted on infrastructure you control, and uses HTTPS with proper authentication. Consider replacing MCP tools with Power Platform custom connectors that provide DLP policy enforcement and governance controls.
Learn more: Use MCP servers in Copilot Studio
Mito check. Author: lnfernux. Open-source licences.
Open-source licences. CIS Benchmark material © Center for Internet Security, Inc.; CISA SCuBA baselines are public domain (CC0).
© 2026 Virtual Petals. Mito is a Virtual Petals service, hosted in Microsoft Azure UAE North.