Microsoft 365 outage disrupts Copilot Chat and Outlook as recovery continues

Microsoft said on November 13 that some Microsoft 365 services were improving after a widespread outage, with capacity still below normal. The company's 10:01 UTC update said Microsoft 365 Copilot users could have trouble opening Copilot Chat or getting replies in the web and desktop apps. Copilot features that rely on work email, calendars and files could also fail.

The University of Pennsylvania’s public feed of Microsoft notices dates the first Exchange Online mailbox problems to 16:17 UTC on November 12. Outlook users could have trouble opening mailboxes or receiving new messages. Microsoft said it had extended the email timeout window to 48 hours to give delayed messages more time to remain in the delivery queue. The company was prioritizing that backlog while it brought affected infrastructure back into service.

Teams calls largely work as calendar features falter

Microsoft traced the scope of the Microsoft 365 disruption to features that rely on Exchange Online, which handles work mail and calendar data. Its notices covered Outlook in the browser and the desktop app, along with other ways of connecting to a mailbox. The Exchange Online advisory also said messages between an on-premises mail system and Exchange Online could be affected. The University of Pennsylvania's public copy of Microsoft's incident feed records the mailbox disruption under EX1550543 and the wider service incident under MO1550821.

Microsoft said people could generally still join Teams calls and meetings, send group chats and post in standard channels. Scheduling a meeting, viewing the calendar or editing an invitation could fail. Teams could also have trouble returning Copilot responses that draw on Outlook, calendars or work files. Files stored directly in SharePoint remained accessible, though some users could run into trouble sharing files with people outside their organization where external sharing was enabled.

The effects varied by organization. The University of Bath said in an 08:30 UTC update that basic Teams functions appeared to work while Outlook was inaccessible to its staff. Kyoto University reported connection problems with Outlook, Teams and OneDrive, including the service that handles student email. The University of Queensland included Exchange Online and Microsoft 365 Copilot in a notice to staff and students. The university directed staff to approved alternative ways to communicate and said exams would proceed as scheduled.

ChatGPT and Claude report connector errors

OpenAI reported problems in the ChatGPT connector for Microsoft 365. It said users might be unable to set up new connections and tools linked through the connector might fail. Its status notice lists ChatGPT's Connectors and Apps as the affected component.

Anthropic attributed failures in its own Microsoft 365 connector to the Microsoft outage. In its 22:23 UTC update on November 12, it said most requests to the connector's mail, calendar and search tools were failing. Claude for Outlook was also affected. An earlier update said some users might be unable to open the Outlook add-in. Anthropic listed a separate incident for sign-in errors on Claude's own services.

Microsoft brings servers back in stages

In updates relayed through the University of Pennsylvania, Microsoft said servers it returned to service could stop responding again once they received traffic. The company tested recovery on a portion of its infrastructure by isolating servers, restarting affected components and gradually sending traffic back. It reported improved stability after applying a patch to part of the system that handles authentication and routing. The rollout then expanded to infrastructure serving regions entering their busiest hours.

Microsoft's Friday update said engineers were still managing traffic and prioritizing the delivery queue as services recovered.

The company said it had found no evidence of malicious activity and would publish its findings on the root cause and steps to prevent a recurrence after mitigation. Its 10:01 UTC notice put the next status update at or before 11:00 UTC on November 13.