For almost a full day, from 16:20 UTC on 12 November 2026 to 16:10 UTC on 13 November, most Microsoft 365 users could not open their mailboxes. Microsoft's preliminary incident report, dated 15 November, puts the blame on a defect in a Windows fix that had been running on its own networking gateway servers since July. Within a day of the outage, phishing emails dressed up as Microsoft outage notices were already landing in inboxes.
What happened
Microsoft tracked the event as EX1550543 (Exchange Online) and MO1550821 (the wider Microsoft 365 impact). Exchange Online took the worst of it. Outlook, Outlook on the web, Exchange ActiveSync and the REST APIs all failed to connect for most users, and incoming mail was delayed.
Features in Teams, Copilot, SharePoint, OneDrive, Defender and Purview that depend on Exchange Online were also affected. Teams, OneDrive, SharePoint and the Office apps were able to move off the failing gateway and came back sooner. Exchange Online could not, because of its scale.
Microsoft has said the outage was a technical failure and not a cyberattack.
Why it happened
The report describes three conditions that lined up on the same day.
- A dormant Windows bug. In July, Microsoft began rolling out a Windows fix built specifically for its Microsoft 365 gateway servers. Under certain loads, lock contention inside an internal Windows process used up CPU and memory until a server stopped responding. The bug is worse on newer machines with more cores, while Microsoft's load tests mostly ran on weaker hardware, so it went unnoticed for more than three months. Microsoft says neither the bug nor the fix affects ordinary Windows installations.
- Extra traffic. At 01:07 UTC on 12 November, about 15 hours before the outage, an automated failover moved Teams traffic onto the gateway that Exchange Online also uses.
- Less capacity. Nearly 14% of that gateway's servers were already offline because of an unrelated problem from the day before.
When the first servers stopped responding, traffic management did what it normally does and sent users to healthy servers. That extra load pushed those servers into the same state. In under two hours, more than 80% of the gateway servers were unresponsive, and a separate issue kept the regional boundaries from containing the failure.
Recovery was slow. Servers brought back online were flooded by queued traffic and went down again, and because the faulty change was months old, the usual rollback tools could not undo it quickly. Engineers built an update that returned the connection-handling code to its earlier working state and applied it to groups of servers starting at 07:20 UTC on 13 November. Global webmail availability was back to normal by 14:17 UTC. Microsoft extended the mail delivery window so that held messages were delivered instead of bounced. A final report is due on 20 November.
The phishing that followed
On 13 November, Microsoft's threat intelligence team flagged emails posing as Microsoft 365 service health notices. They cite the real incident number EX1550543 and claim that mail held during the outage will be deleted within 24 hours unless the recipient clicks a link to "restore" it.
Signs to look for:
- subject lines starting with "Outage -" followed by a six-character code;
- sender names ending in "Outage Notification" or "Service Health";
- sending domains moments-notice[.]com, planetmaison[.]com and newbooker[.]com;
- SendGrid tracking links that redirect to reallifeaccounting[.]com.
Microsoft describes the campaign as consistent with credential phishing and says Defender is blocking it. Whether any passwords were actually captured has not been confirmed. Organisations in several US sectors were targeted.
What to do now
- Treat outage emails as suspect. Microsoft's advice is to check incident status only in the Microsoft 365 admin center, not through links in an email; Microsoft explains where to find them in its service health guide.
- Search for the campaign. Look in mail logs and quarantine for the subject pattern and the three sender domains above, and check whether anyone clicked.
- Review Safe Links. Make sure time-of-click checks are on for email, Teams and Office apps, and that users cannot click through warning pages. The settings are covered in Microsoft's Safe Links documentation.
- Reset and review if someone clicked. Reset the password, revoke active sessions and check sign-in logs and inbox rules for that account.
- Plan for a mail outage. Decide in advance how staff will be told about an incident when email itself is down, so a fake "status update" has less room to work.
If a user did enter a password on one of these pages, the question is where it ends up next. Our Deep/Dark Web monitoring watches leak sites and criminal markets for your company's credentials, so a stolen login can be spotted and reset before it is used.
