Ai Agent Detection Protection
In brief
The documentation now directs administrators to configure prompt evidence collection for AI agent alerts and links local Windows agents to a device-level setting. Author metadata was also updated.
What Defender admins need to know
Review the prompt evidence setting if alert evidence should include triggering prompt snippets; no required action is stated.
Summaries are generated from the documentation change itself.
Documentation change
The comparison below shows only the changed extract. Use the full-page view for complete context.
Threat detection is supported only for published Microsoft Foundry agents. Unpublished agents, including agents used only in a playground environment, aren’t supported. For more information, see [Publish agents to Microsoft Copilot and Microsoft Teams in the Foundry portal](/en-us/azure/foundry/agents/how-to/publish-copilot?tabs=portal)
- To detect threats to local AI agents that run on endpoints, set up AI agent runtime protection in Microsoft Defender for Endpoint. Microsoft Defender for Endpoint must run in active mode. Local agents are onboarded separately from cloud agents.
- (Optional)
To includeConfigure prompt evidence collection for AI agent alerts. For local Windows agents, theprompt snippets that triggered a detection as evidence in alerts, enable prompt evidence collection. This settingdevice-level setting enables or disables collection when AI agent runtime protection isenabled by default.configured. See the overview for feature defaults and privacy guidance.
Detect AI agent threats in near-real-time
@@ -9,8 +9,8 @@ feedback_product_url: https://techcommunity.microsoft.com/t5/microsoft-365-defen uhfHeaderId: MSDocsHeader-MicrosoftDefender manager: orspodek description: Learn how to detect, investigate, and hunt for threats to AI agents using Microsoft Defender.-ms.author: guywild-author: guywi-ms+ms.author: joshbregman+author: j0shbregman ms.reviewer: itaicohen ms.service: microsoft-defender ms.update-cycle: 180-days@@ -66,7 +66,7 @@ Microsoft Defender detects and enables security teams to investigate threats to
Threat detection is supported only for published Microsoft Foundry agents. Unpublished agents, including agents used only in a playground environment, aren’t supported. For more information, see [Publish agents to Microsoft Copilot and Microsoft Teams in the Foundry portal](/en-us/azure/foundry/agents/how-to/publish-copilot?tabs=portal)
- To detect threats to local AI agents that run on endpoints, set up [AI agent runtime protection in Microsoft Defender for Endpoint](/en-us/defender-endpoint/configure-ai-agent-runtime-protection). Microsoft Defender for Endpoint must run in active mode. Local agents are onboarded separately from cloud agents.
-- *(Optional)* To include the prompt snippets that triggered a detection as evidence in alerts, enable [prompt evidence collection](ai-agent-real-time-protection#control-prompt-evidence-in-alerts). This setting is enabled by default.
+- *(Optional)* Configure [prompt evidence collection](ai-agent-real-time-protection#control-prompt-evidence-in-alerts) for AI agent alerts. For local Windows agents, the [device-level setting](/en-us/defender-endpoint/configure-ai-agent-runtime-protection#configure-prompt-evidence-collection) enables or disables collection when AI agent runtime protection is configured. See the [overview](/en-us/defender-endpoint/ai-agent-runtime-protection-overview#prompt-evidence-collection) for feature defaults and privacy guidance.
## Detect AI agent threats in near-real-time
