Ai Agent Runtime Protection Overview
In brief
The documentation adds prompt evidence collection for AI agent runtime protection, including its Audit/Block prerequisites, alert behavior, upcoming default enablement, privacy considerations, and PowerShell or policy-template configuration.
What Defender admins need to know
Review privacy and data-handling requirements and consider explicitly configuring the setting; explicit configurations override the upcoming default.
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.
For local PowerShell and organization-wide Microsoft Intune configuration options, see Set up AI agent runtime protection with Microsoft Defender for Endpoint.
Prompt evidence collection
Prompt evidence collection is a device setting for AI agent runtime protection. When collection is enabled and runtime protection is enabled in Audit or Block mode, Defender exposes prompt snippets exchanged between the user and the agent to support deeper analysis of agent-related alerts. These snippets include only the portions of user prompts or agent responses identified as suspicious and relevant to security classification.
Note
An upcoming Microsoft Defender Antivirus platform release will enable prompt evidence collection by default. Devices that haven't received the update that changes this default still have collection disabled by default. Explicitly configured settings take precedence over this default. This change doesn't enable runtime protection or change its audit or block mode.
Important
Sensitive data and secrets are redacted from prompt snippets shown in alerts, but customer conversations might still be sensitive in nature. Review your organization's privacy and data-handling requirements before you enable collection. Evidence included in security alerts is retained according to the Microsoft Defender XDR data retention policy.
Disabling device-level collection doesn't disable runtime protection. Content associated with activity that Defender blocks can still appear in the associated security alert.
Configure device collection through PowerShell or the policy template. For prerequisites, commands, and deployment steps, see Configure prompt evidence collection.
Known issues (Preview)
Important
@@ -14,7 +14,7 @@ ms.author: joshbregman ms.service: defender-endpoint ms.topic: overview ms.custom: msecd-doc-authoring-1030-ms.date: 2026-10-08T00:00:00.0000000Z+ms.date: 2026-11-11T00:00:00.0000000Z ai-usage: ai-assisted locale: en-us document_id: 2840101c-b02b-c6b4-ccda-603c72dcf7d6@@ -102,6 +102,22 @@ Microsoft recommends starting in audit mode to observe detections and validate a
For local PowerShell and organization-wide Microsoft Intune configuration options, see [Set up AI agent runtime protection with Microsoft Defender for Endpoint](configure-ai-agent-runtime-protection).
+### Prompt evidence collection
+
+Prompt evidence collection is a device setting for AI agent runtime protection. When collection is enabled and runtime protection is enabled in **Audit** or **Block** mode, Defender exposes prompt snippets exchanged between the user and the agent to support deeper analysis of agent-related alerts. These snippets include only the portions of user prompts or agent responses identified as suspicious and relevant to security classification.
+
+Note
+
+An upcoming Microsoft Defender Antivirus platform release will enable prompt evidence collection by default. Devices that haven't received the update that changes this default still have collection disabled by default. Explicitly configured settings take precedence over this default. This change doesn't enable runtime protection or change its audit or block mode.
+
+Important
+
+Sensitive data and secrets are redacted from prompt snippets shown in alerts, but customer conversations might still be sensitive in nature. Review your organization's privacy and data-handling requirements before you enable collection. Evidence included in security alerts is retained according to the [Microsoft Defender XDR data retention policy](/en-us/defender-xdr/data-privacy).
+
+Disabling device-level collection doesn't disable runtime protection. Content associated with activity that Defender blocks can still appear in the associated security alert.
+
+Configure device collection through PowerShell or the policy template. For prerequisites, commands, and deployment steps, see [Configure prompt evidence collection](configure-ai-agent-runtime-protection#configure-prompt-evidence-collection).
+
## Known issues (Preview)
Important
