It wasn't a Copilot-only failure. Copilot sits on top of Exchange Online, Microsoft Graph and the rest of Microsoft 365, so when Exchange went down, Copilot had both its own problems and nothing to read.
The short answer
- Microsoft 365 Copilot (work/school accounts): yes, affected. Microsoft named it in MO1550821.
- Consumer Copilot on a personal Microsoft account: unconfirmed. Microsoft's advisory names the work product, and Nacionale, which first broke out the Copilot angle, said none of the reports it reviewed settled the question.
- Root problem: Exchange Online, tracked separately. Exchange Online mailbox connectivity is also being tracked separately under EX1550543.
- Cyberattack? No evidence of one. Microsoft's public updates describe infrastructure failures and recovery work, rather than presenting evidence of a confirmed cyberattack behind this disruption.
What "Copilot is affected" actually means
Microsoft posted status updates at 00:00 UTC and 01:12 UTC on November 13. Both said users might not be able to load Copilot Chat or get answers in the web and desktop apps. They also said Copilot might not be able to reach work data such as email, calendar and files. IT-Connect put it plainly: the chat may fail to load or may no longer access emails, calendars and files.
The word "may" matters. Microsoft's updates also said features that interact with Exchange Online and email were affected, while parts that don't depend on Exchange kept working. Some users got errors on every prompt. Others could still use Copilot for general questions but got nothing when they asked about their own inbox.
The Teams section of the status record has the most specific Copilot detail. It lists failures with Copilot answers about Outlook, calendar or Microsoft 365 files, plus problems reading history and adding Copilot to new chats. It also says calls, meetings, group chats and channel messages mostly kept working.
Why an email outage takes Copilot down with it
This follows from how Microsoft licenses and builds Copilot. Microsoft's documentation splits Copilot into tiers:
| Tier | Where answers come from | How exposed it was to this outage |
|---|---|---|
| Copilot Chat (Basic) | Mainly the public web; you supply work content yourself | Less dependent on Exchange, but the chat itself might not load |
| Microsoft 365 Copilot (Basic) | Web data plus in-app features in Word, Excel, PowerPoint, OneNote | Outlook features need a working mailbox |
| Microsoft 365 Copilot (Premium) | Web data plus emails, files, meetings and calendars, pulled automatically through Microsoft Graph | Most exposed: grounding in work data depends on the failing services |
Premium is the paid add-on, and it's the one that can answer "summarize my unread mail" or "what's on my calendar tomorrow." Those answers come straight from Exchange-backed data. If the mailbox can't be reached, Copilot has nothing to summarize.
Copilot inside Outlook has the same problem. Microsoft lists drafting email and summarizing threads among its Outlook features, but none of that works if the mailbox won't open.
Copilot Chat Basic looks like a fallback because it mostly uses the web. But Microsoft warned the chat itself might not load, so it wasn't a reliable backup during the outage.
Bottom line: the more Copilot depends on your organization's data, the harder this outage hit it.
What went wrong underneath
Microsoft's explanation changed several times overnight. Times below are UTC, with Eastern Time in parentheses where useful:
- 16:17 UTC, Nov 12: Exchange Online incident EX1550543 starts, covering mailbox connection failures.
- 18:26 UTC (1:26 p.m. ET): the broader incident MO1550821 opens. Microsoft first blamed elevated load across multiple regions.
- 19:30 UTC: Microsoft corrects itself. In its words, "We determined the elevated load isn't driven by an increase in traffic, but by a number of servers being unexpectedly unavailable; straining the remaining infrastructure." All deployments and configuration changes are paused.
- 19:43 to 20:57 UTC: Microsoft finds servers failing automated health checks, suspects a recent change to a health-check component, and tries a different health-check method that brings some servers back.
- About 22:00 UTC: Microsoft says its fixes haven't brought the expected relief. A core process on the affected servers is overwhelmed, causing them to stop responding.
- 22:58 UTC: Microsoft reports limited recovery in some regions after returning some servers to service.
- Overnight: Microsoft says it is starting to revert a recent network configuration change that lines up with when the problems began. It also identifies a process that handles secure connections as one of the parts getting overwhelmed.
- 08:06 UTC, Nov 13: the EX1550543 record still lists the service as degraded. Microsoft says it is still rolling out a patch step by step and will publish a post-incident report after mitigation.
Astra Security's analysis described the failure as a cascade: those remaining servers then saturated and began failing their own checks, creating a feedback loop that chewed through capacity across the region. That fits Microsoft's own updates, but it's an outside interpretation, not Microsoft's root-cause report.
Reports disagree on when the outage ended. Astra's summary says the outage lasted roughly 10 hours, from about 17:54 UTC to 05:33 UTC the next morning. But Microsoft's own EX1550543 entry at 08:06 UTC still listed Exchange Online as degraded. Treat any "it's over" claim with caution until your tenant's service health page says the incident is resolved.
The wider damage
Copilot was one of many affected services. Microsoft's updated impact list included:
- Teams: Exchange-dependent scenarios such as calendars and meeting scheduling. Core messaging and meetings mostly worked.
- SharePoint Online: features that depend on Exchange. Files stored directly in SharePoint stayed accessible.
- Purview: problems with DLP, auto-labeling, eDiscovery search and export, and loading the portal.
- To Do: the app might not load at all.
- Power BI: exporting reports with MIP sensitivity labels to PDF could fail.
- Windows 365: Cloud PCs set up to use Purview Customer Key might not connect.
- Also listed: Defender XDR portal pages, Intune, Planner, Defender for Office 365 Safe Links and Universal Print.
Some admins also had trouble reaching the place they would normally check. Astra noted that the admin center showed loading issues on some pages, under incident MO1550821, which meant some administrators could not reach Microsoft's own health information during the event.
Which Copilot are you using?
Since only the work product is confirmed as affected, it helps to know which one you're signed in to. Microsoft's Copilot overview says work and school sessions show a different background color, a "Work" label under the user profile in the navigation pane, and a green data protection shield. If you don't see those, you're probably on a personal Microsoft account. Microsoft hasn't said whether that version was affected.
Downdetector doesn't answer the question either. According to Downdetector, which tracks user-submitted outage reports, complaints surged Thursday afternoon across numerous Microsoft services, including Outlook, Microsoft 365, Teams, Copilot, Azure, OneDrive, and the Microsoft Store. Those counts don't separate work accounts from personal ones.
What to do while it's broken
For users:
- Don't keep signing out and back in. The fault is on Microsoft's servers, and re-authenticating won't fix it. Nacionale also noted that OpenAI's status page warned new Microsoft connections might fail during the incident.
- Paste content in yourself. If an email or document is still visible, for example in a cached copy, paste the text into an assistant that doesn't need your Microsoft account. Check your organization's data policy before putting work content into a third-party tool.
- Expect partial recovery. Microsoft said recovery could be gradual as capacity returns and apps reconnect. Copilot may come back before or after Outlook, since it has to reconnect to mail and calendar data as well.
For admins:
- Sign in to the Microsoft 365 admin center and go to Health > Service Health, then search for MO1550821.
- Track EX1550543 too, since that's where the underlying Exchange fix is reported.
- If the admin center won't load, Microsoft says to use status.cloud.microsoft if the admin center is not available to you.
- Tell users plainly which Copilot features rely on Exchange so the helpdesk doesn't get flooded with tickets about broken calendar summaries.
- Watch for the post-incident report. Microsoft has committed to one after mitigation, and it should be the authoritative explanation of the cause.
The bigger lesson
Microsoft sells Microsoft 365 Copilot on the strength of its access to your email, files and meetings through Microsoft Graph. This outage showed the other side of that: when Exchange goes down, the assistant built on it goes down too. It's hard to call that a design flaw, because an assistant grounded in your mail can't work without your mail. But it's a real resilience question for organizations that now rely on Copilot for routine work.
Expect plenty of business-continuity reviews that previously treated "AI assistant" and "email" as separate items. After this week, it makes more sense to plan for them as one dependency.
Microsoft hasn't given a full restoration time or said whether personal-account Copilot was touched, and it hasn't published a final root cause. Until it does, check your tenant's service health page rather than social media.