MC1422060 High
(Updated) Microsoft Defender for Office 365: Prompt injection protection for email
Summary AI-generated
Prompt injection protection for email in Defender for Office 365 detects and blocks malicious prompt injection content, quarantining threats before AI workflows process them.
Suggested actions AI-generated
- Review submission and quarantine workflows
- Use Defender submission process for false positives
- Use Tenant Allow Block List for exceptions
- Inform security and helpdesk teams about the new detection category
Written by Azure OpenAI (gpt-4.1) from the text of the post below. It can be incomplete or wrong; the original post is authoritative.
Similar posts
Search for more like this- MC1166867 (Updated) Microsoft Defender for Office 365: Enhancing the quarantine email preview experience
- MC1476237 Microsoft Defender for Office 365: Remediation actions from the Teams message entity flyout
- MC1440701 MDO Encrypted email attachment protection
- MC1163754 Enhancements to the Deep Analysis tab of Email Entity page by Microsoft Defender for Office 365
- MC1279093 (Updated) Microsoft Defender for Office 365: Enhancing how we handle promotional mail
- MC1171845 (Updated) Microsoft Defender for Office 365: Enhancing the quarantine experience for administrators
Original post from Microsoft
Updated September 2, 2026: We have updated the timeline. Thank you for your patience.
[What and Why]
We are introducing prompt injection protection for email in Microsoft Defender for Office 365. This capability detects and blocks malicious prompt injection content embedded in email messages that attempt to manipulate AI assistants and agents. It helps protect enterprise data by identifying attacks designed to exfiltrate information, discover tools, or expose system prompts. High confidence threats are automatically quarantined before they can be processed by AI powered workflows. This enhancement strengthens enterprise ready AI security and aligns with evolving threat patterns.
[Rollout Schedule]
- Public Preview: Beginning early July 2026 and expected to complete by early September 2026
- General Availability (Worldwide): Beginning early October 2026 (previously early September) and expected to complete by early October 2026 (previously early September)
[Impact on Your Organization]
Who is affected
- Organizations with Microsoft Defender for Office 365 Plan 2 or Microsoft 365 E5
Platforms and services
- Exchange Online, Microsoft Defender for Office 365, Microsoft Defender XDR services
What will happen
- Emails identified as prompt injection will be classified as High Confidence Phish.
- A new Detection Technology value called Prompt Injection Protection will be applied.
- High confidence threats will be automatically quarantined.
- The feature is enabled by default for eligible tenants.
- Existing policies and workflows remain unchanged.
- These detections will appear within existing threat investigation and reporting experiences in Microsoft Defender.
[Action Required / Recommendations]
No action is required.
Recommended actions:
- Review your submission and quarantine workflows.
- Use the Microsoft Defender submission process if false positives occur.
- Use Tenant Allow Block List if needed to manage exceptions.
- Inform your security and helpdesk teams about the new detection category.
[Compliance considerations]
No compliance considerations identified, review as appropriate for your organization.
Additional details from Microsoft
- Summary
- Microsoft Defender for Office 365 will introduce prompt injection protection for email, detecting and quarantining malicious content targeting AI assistants. Rolling out from July to October 2026, it affects organizations with Defender Plan 2 or Microsoft 365 E5, enabling automatic high-confidence threat quarantine by default with no required action.