Forum Discussion
Solved
Defender for AI data storage/processing
Hi, does anyone know where the data that Defender for AI uses is processed and what data is stored and available to Microsoft?
If abuse monitoring is turned off, the documentation says "Microsoft does not store the prompts and completions associated with the approved Azure subscription."
If content filtering is enabled the documentation says "Noo prompts or generated content are stored in the content classifier models."
But I was wondering what data is stored/processed, where this happens and if there's any documentation around this for the Defender for AI service.
Could anyone point me to a page, please?
Thanks,
Neil.
Hey Neil,
When Defender for AI is enabled, it does collect and process certain data—primarily prompts and model responses if the “User Prompt Evidence” feature is turned on. This data is stored securely in your selected Azure region (via Log Analytics) and is used strictly for security monitoring, alerting, and abuse detection. Microsoft explicitly states that this data is not used to train or improve models, nor is it shared with OpenAI or other customers. Defender for AI may also collect additional metadata, such as configuration information and event logs, to enhance its threat detection capabilities. If you integrate with Microsoft Purview, prompt and response data may be further analyzed for compliance purposes, but all of this stays within your control. Your Log Analytics settings govern retention and storage. I hope this helps!
3 Replies
Microsoft stores collected data in its global network of data centers, which are strategically located across different regions to ensure security, compliance, and performance. The exact location depends on the service being used and the user's geographic region.
For example:
Microsoft 365 data is stored in regional data centers based on the user's location and compliance requirements.
Azure data is stored in Microsoft’s cloud infrastructure, with options for customers to select specific regions.
Defender for AI processes data within Microsoft's cloud infrastructure, ensuring security and compliance.
thanks Vimal. What I'm looking for, specifically, is what additional information is stored/processed by Microsoft as a result of enabling Defender for AI service? The answer might be nothing, but I'm struggling to find any evidence of this.
Thanks,
Neil.
Hey Neil,
When Defender for AI is enabled, it does collect and process certain data—primarily prompts and model responses if the “User Prompt Evidence” feature is turned on. This data is stored securely in your selected Azure region (via Log Analytics) and is used strictly for security monitoring, alerting, and abuse detection. Microsoft explicitly states that this data is not used to train or improve models, nor is it shared with OpenAI or other customers. Defender for AI may also collect additional metadata, such as configuration information and event logs, to enhance its threat detection capabilities. If you integrate with Microsoft Purview, prompt and response data may be further analyzed for compliance purposes, but all of this stays within your control. Your Log Analytics settings govern retention and storage. I hope this helps!
