Forum Discussion
Microsoft Defender AI agent runtime protection: the series guide (parts 1 to 5)
Hi all,
https://cloud-secure.ai/2026/10/26/microsoft-defender-ai-agent-runtime-protection-the-series-guide-parts-1-to-5/
I’ve created a five-part blog series that I hope will provide useful insights and practical guidance for the community. I’d really welcome your thoughts, feedback and experiences, and hopefully the series can help others on their adoption journey.
Throughout the series, Contoso is used as the fictional organisation across all scenarios and examples.
AI agents such as Claude Code, GitHub Copilot and Gemini CLI are already on managed endpoints, running with the full rights of the signed-in user and connecting to MCP servers that nobody has reviewed. This series works through what Microsoft Defender for Endpoint can do about that today – from detecting prompt injection, to finding the agents, to controlling and automatically blocking the servers they talk to.
Every part is built and tested in a lab tenant, with screenshots of what actually happened – including where the product or the documentation did not behave as expected.
I have created a 5 part blog series , I would welcome any comments and hope this series can help the community. The demo company in all posts is Contoso.