Microsoft 365 outage breaks Claude's connector; Claude sign-in fails too

Microsoft 365 was still degraded at 07:25 UTC on Friday 13 November 2026, more than 15 hours after users lost access to Exchange Online mailboxes at 16:17 UTC on Thursday, according to Microsoft's service-health updates as quoted by [CIO](https://www.cio.com/article/4248249/microsoft-struggles-to-get-m365-services-back-online-after-widespread-outage.html) and [CX Today](https://cxtoday.com/microsoft-outage-cut-agents-off-from-the-contact-center). Microsoft gave no time for full recovery and said overnight that its fixes were "not taking as expected", CIO reports. Teams calls and meetings have mostly kept working; calendars and scheduling have not.

The AI assistants built on Microsoft 365 stopped with the mailboxes. Microsoft's notice says Microsoft 365 Copilot users may be unable to load Copilot Chat, and that Copilot may be unable to reach email, calendars and files, CIO reports. Anthropic opened [an incident on its status page](https://status.claude.com/incidents/sqn3fg1l96dl) at 17:52 UTC on Thursday and wrote at 22:23 UTC: "This issue is caused by a Microsoft 365 service incident, and mail, calendar and search tools in the Microsoft 365 connector are still failing for most requests. Claude for Outlook remains affected." The connector is the link that lets Claude read a user's Outlook mail, calendar and SharePoint files. That incident was still open at 10:58 UTC on Friday, 17 hours on. CX Today reports that OpenAI's status page showed an open issue with its Microsoft connectors at 07:37 UTC.

A second Claude incident, with no stated link to Microsoft

At 09:58 UTC on Friday Anthropic opened [a second incident](https://status.claude.com/incidents/nn11wk485zfn), rated major: "We are investigating an issue with users authenticating to Claude.ai and Claude Code." Its status page marks claude.ai as a major outage and the Claude Console and Claude Code as partial outages. Anthropic gives no cause, does not connect it to Microsoft's outage, and had posted no further update an hour later.

Support: the contact centre went down for eight hours

Dynamics 365 Contact Center, Microsoft's contact-centre product, was down from 17:00 UTC on Thursday until 00:57 UTC on Friday, CX Today reports from Microsoft's incident CR1550885. Agents who could not sign in could take neither calls nor messages. Microsoft's preliminary cause is separate from the mail failure: "A portion of the database infrastructure responsible for Contact Center experienced connectivity failures." Microsoft also says Power Platform flows and Copilot Studio connectors are failing intermittently, according to CX Today.

A security control went with it

Microsoft's notice says Defender for Office 365 Safe Links, which checks a link when someone clicks it, may be unable to verify URLs clicked in Outlook and Teams, CIO reports. Sanchit Vir Gogia, chief analyst at Greyhound Research, told CIO that companies need "a written decision on whether each security control fails open or closed": whether an unchecked link is let through or blocked.

What is not known

Microsoft has published no final account of the cause. It said a process on affected servers "responsible for establishing secure connections becomes overwhelmed, causing the servers to stop processing requests", CIO reports. At 07:25 UTC it reported "positive results for server stability" after patching part of its infrastructure, according to CX Today. On an attack, its wording moved from "no indication this is the result of malicious activity" at 04:11 UTC to "As of now, we are not aware of any malicious activity that is causing this incident" at 05:46 UTC, [The Register](https://www.theregister.com/saas/2026/11/13/microsofts-massive-m365-outage-stretches-past-12-hours-as-its-fixes-keep-failing/5309609) reports. No source says how many organisations are affected, what service credits Microsoft will offer, or what the lost hours cost.

What to do today

• IT: tell staff now that Copilot, Claude's Microsoft 365 connector, Claude for Outlook and claude.ai sign-in are affected, so they stop retrying and stop filing tickets. Keep a log of when each service dropped and returned and how many users were hit, under Microsoft incidents EX1550543 and MO1550821, and ask your licensing contact how and by when to claim a service credit.
• IT and security: ask your Microsoft 365 administrator what Safe Links does with a link it cannot check, and warn staff about links in email and Teams until Microsoft confirms it is working.
• Support: if you run on Dynamics 365 Contact Center, record the eight hours and what customers saw, then test the one channel customers can still reach when the contact centre, email and your AI assistant fail together.