What happened
As part of the June 30, 2026 Microsoft Security update, Microsoft Defender now auto-discovers over 25 types of local AI agents and MCP servers on endpoints. The capability adds runtime protection against prompt injection attacks targeting developer coding agents including GitHub Copilot CLI and Claude Code, extending endpoint protection policies to cover the local AI execution plane.
Why it matters
Local AI agents and MCP servers represent a largely unmonitored attack surface on developer workstations. Defender's ability to discover and protect these at runtime is the first mainstream EDR capability to cover this class of threat, directly countering the prompt-injection and tool-poisoning attack patterns disclosed in the same week.
Applicability
Any enterprise with Defender for Endpoint deployed and developers using AI coding agents (GitHub Copilot, Claude Code, etc.); evaluate now and ensure agent discovery is enabled in Defender settings.