Microsoft Defender for Cloud

AI model security (Preview)

4 September 2026Doc update+1 / −1 lines

In brief

The documentation reiterates that AI model scanning is in preview, included with the Defender for AI Services plan, and has no additional scanning charge during preview; related threat protection costs may still apply.

What Defender admins need to know

Review the updated cost guidance when evaluating or budgeting for AI model scanning. No administrator action is stated.

Summaries are generated from the documentation change itself.

Documentation change

View full page diff

View on Microsoft Learn

The comparison below shows only the changed extract. Use the full-page view for complete context.

AI model security (Preview)

Important

This feature is currently in preview and included with the Microsoft Defender for AI Services plan. During preview, there is no additional charge for AI model scanning. However, enabling the Defender for AI Services plan may incur costs related to threat protection features. Continued inclusion of AI model scanning feature as part of Defender for AI Services is not guaranteed when it becomes generally available (GA), and licensing requirements may change. If that occurs, a notification will be sent before the feature is disabled with options to re‑enable it under the applicable license. AI model scanning is intended to assist customers in identifying potential security risks within supported model artifacts. Scan results may not identify all malicious, unsafe, or otherwise abusive content and may produce false positives or false negatives. Detection coverage varies based on supported model formats, scanning techniques, and available threat intelligence. Customers should independently review and validate scan findings and should not rely solely on scan results when making deployment, security, or compliance decisions.

As organizations increasingly use artificial intelligence (AI) models to drive automation, insights, and intelligent decision-making, security teams need visibility and control to assess the safety and compliance of AI models entering their environments. These models often have broad access to data and infrastructure. Without these capabilities, it becomes increasingly difficult to enforce internal standards.

@@ -10,7 +10,7 @@ ai-usage: ai-assisted # AI model security (Preview)  > [!IMPORTANT]-> This feature is currently in preview and included with the Microsoft Defender for AI Services plan. During preview, there is no additional charge for AI model scanning. However, enabling the Defender for AI Services plan may incur costs related to threat protection features. Continued inclusion of AI model scanning feature as part of Defender for AI Services is not guaranteed when it becomes generally available (GA), and licensing requirements may change. If that occurs, a notification will be sent before the feature is disabled with options to re‑enable it under the applicable license.+> This feature is currently in preview and included with the Microsoft Defender for AI Services plan. During preview, there is no additional charge for AI model scanning. However, enabling the Defender for AI Services plan may incur costs related to threat protection features. Continued inclusion of AI model scanning feature as part of Defender for AI Services is not guaranteed when it becomes generally available (GA), and licensing requirements may change. If that occurs, a notification will be sent before the feature is disabled with options to re‑enable it under the applicable license. AI model scanning is intended to assist customers in identifying potential security risks within supported model artifacts. Scan results may not identify all malicious, unsafe, or otherwise abusive content and may produce false positives or false negatives. Detection coverage varies based on supported model formats, scanning techniques, and available threat intelligence. Customers should independently review and validate scan findings and should not rely solely on scan results when making deployment, security, or compliance decisions.  As organizations increasingly use artificial intelligence (AI) models to drive automation, insights, and intelligent decision-making, security teams need visibility and control to assess the safety and compliance of AI models entering their environments. These models often have broad access to data and infrastructure. Without these capabilities, it becomes increasingly difficult to enforce internal standards.