Microsoft 365 Outage November 2026: What Happened and Was AI Involved?
Key Takeaways
- A global Microsoft 365 outage began on Thursday 12 November 2026, disrupting Outlook, Teams, SharePoint, Copilot and other services, with outage reports coming from 74 countries and territories.
- Microsoft traced the disruption to a fault introduced during an earlier network capacity upgrade, which surfaced after a more recent network change. Servers failed automated health checks and dropped offline, and a core process handling secure connections became overwhelmed as traffic shifted to the remaining servers.
- Microsoft has said explicitly that the outage was not related to cyber activity or to AI and AI agents. There is no evidence that AI caused the outage, and Microsoft has ruled out a cyberattack.
- The disruption lasted about a day. Microsoft reported broad recovery by Friday 13 November, with access largely restored, but the incidents were still open at the end of 13 November and a full post-incident report has been promised.
- The search query “is microsoft still down” reached number 10 in the UK Trending Now list, with more than 500,000 searches, showing how widely the outage was felt.
- This was the third major Microsoft 365 disruption of 2026, following multi-hour outages in January and a code-update failure in August that took several days to fully resolve.
What Went Down, And When
The outage began on Thursday 12 November 2026. Microsoft logged the Exchange Online incident in the late afternoon UTC, under the incident reference EX1550543, and the wider Microsoft 365 incident under MO1550821. Reports to the outage-tracking service Downdetector about Outlook began climbing at about 17:25 UTC, and within about an hour outage reports were coming from 74 countries and territories. The speed of the spread is one of the striking facts about this incident: it was global within an hour, not a regional failure that grew.
The services affected were the core of Microsoft’s workplace cloud. Exchange Online, Microsoft’s cloud email service, was at the centre of the disruption. Outlook was the most visible casualty, with users unable to reach their mailboxes. Microsoft Teams was disrupted, with problems affecting calendar access and meeting scheduling. SharePoint Online was affected. Microsoft 365 Copilot was affected. Microsoft Defender XDR was affected. The list also included Planner, To Do, Intune and Defender for Office 365. In practice, that meant the email, the meetings, the documents, the AI assistant and the security tools that millions of workplaces depend on were all degraded at the same time.
The disruption was not a brief blip. It lasted about a day, and Microsoft’s own account puts the shape of it clearly. The incident was opened on the afternoon of 12 November and had not been closed by the evening of Friday 13 November, more than 24 hours later. Access was reported largely restored by about 05:33 UTC on 13 November, roughly 11 to 12 hours after the first reports, but Microsoft was still restoring service for remaining customers and regions after that, and it had not declared the incidents over. The distinction between largely restored and fully resolved matters: for most users the outage was over by Friday morning, but for some it continued into Friday and beyond.
The scale of the user impact is visible in the search data. The query “is microsoft still down” reached number 10 in the UK Trending Now list, with more than 500,000 searches. That is a measure of confusion as much as of impact: people were not only affected, they did not know whether the problem was theirs or Microsoft’s, and they were searching for an answer. The same pattern played out in workplaces around the world, where IT teams fielded the same question from every employee at once: is it just me, or is it everyone.
What Microsoft Says Caused It
Microsoft’s account of the cause has developed over the course of the incident, and the most recent version is the one to use. The company says the incident was introduced during an earlier network capacity upgrade, and that it surfaced after a more recent network change. The fault was in the network gateway and routing infrastructure: a recent network change exposed a defect dating from the earlier gateway capacity upgrade, making the infrastructure unstable and taking servers out of service.
The mechanism, as Microsoft has described it, was a cascading failure. A large pool of servers unexpectedly failed automated health diagnostics and dropped offline. As user traffic automatically shifted onto the remaining active servers, a core process on those servers, the process responsible for establishing secure connections, became overwhelmed by elevated processor usage. The overloaded servers then stopped processing requests, reducing available capacity further and shifting still more traffic onto the remaining infrastructure. The result was a domino effect: healthy servers turned unresponsive because they were absorbing the load of the servers that had already failed.
Microsoft’s engineers responded by testing a configuration change to address the problem, restarting affected servers, freezing all routine software deployments and configuration updates, rerouting user traffic onto alternative infrastructure pathways, and deploying an alternative health-check method so that restarted servers could safely rejoin the network. The company also extended some mail-processing timeouts to 48 hours to reduce the risk of queued messages expiring while its systems caught up. By the evening of 12 November, complaint volumes on Downdetector had dropped from their afternoon highs to roughly 32,000 reports, and Microsoft confirmed that recovery was under way, while cautioning that full restoration would remain gradual.
There is one further detail from the reporting that adds context. Tom Warren of The Verge reported that his sources link the outage to configuration changes in Azure Virtual Network, Microsoft’s cloud networking service. That is a source-based report, not Microsoft’s own statement, and it is consistent with Microsoft’s account of a network change exposing an earlier defect. It does not contradict the official account; it adds a specific location for the change.
What Microsoft has not yet done is publish a full post-incident report. The company has promised one, and the detailed root cause, including exactly which change exposed which defect, will only be known when that report appears. Until then, the account above is the best available: a fault introduced during a network capacity upgrade, exposed by a later network change, producing a cascading health-check failure. It is a technical explanation, not a complete one, and the full report may revise the details.
Was AI Involved? No.
The question in the title of this article deserves a direct answer, because it was asked widely and loudly while the outage was happening. Was AI involved in the Microsoft 365 outage? According to Microsoft, no. The company has said explicitly that the outage was not related to cyber activity or to AI and AI agents. A company spokesperson told the ABC there was no indication of a cyberattack, and Microsoft’s executive vice president for Copilot, Agents and Platform, Charles Lamanna, apologised to customers and said there was no indication that the outage resulted from malicious or cyber activity.
The question arose for understandable reasons. The outage hit Copilot, Microsoft’s AI assistant, alongside Outlook and Teams, and it happened on a day when Microsoft was visibly pushing AI to its users. That same day, Microsoft asked users for AI tips, and the top reply was, in effect, make sure your product does not fail. The timing made the AI question feel natural: the company that was telling everyone to use more AI could not keep its own services running. Our earlier report on the Microsoft staff response to the AI push covers that exchange, and the frustration it captured was real.
But frustration is not evidence, and the timing is not a cause. The outage affected services that have nothing to do with AI agents, including basic email delivery and calendar access, and the failure mechanism Microsoft described is a network infrastructure fault, not an AI system failure. No public evidence reviewed establishes that AI agents, Copilot, or any AI system played a role in causing the outage. The speculation that AI caused it, including search queries for “microsoft outage ai attack” and “microsoft outage ai related”, was speculation only, and it has been answered directly by Microsoft.
There is a broader point worth making. When a major cloud service fails, the first theories are often the most dramatic ones: a cyberattack, a sabotage, a rogue AI. The reality is usually more mundane, and this outage is a case in point. A network capacity upgrade introduced a defect. A later change exposed it. Servers failed their health checks. Traffic cascaded onto the survivors. The survivors fell over. That is a story about infrastructure engineering, not about AI, and it is a reminder that the most dangerous failures in cloud computing are often the ordinary ones.
It is also worth being precise about what Microsoft has and has not ruled out. The company has said there is no indication of a cyberattack, no indication of malicious activity, and no connection to AI or AI agents. It has not said that no customer data was affected in any way, and reporting notes that no public evidence reviewed establishes that customer emails were stolen. Those are different claims, and the absence of evidence of theft is not the same as a guarantee that nothing was exposed. The post-incident report should address the data question, and until it does, the careful statement is that Microsoft has found no evidence of a cyberattack or of AI involvement, and no public evidence of data theft has emerged.
The Human Cost Of A Day Without Email
The technical account does not capture what the outage felt like for the people affected, and the human cost is worth describing plainly. For millions of workers, Outlook is not one application among many; it is the place where work arrives, where meetings are scheduled, and where decisions are recorded. When Outlook goes down, work does not stop, but it becomes disorganised, delayed and uncertain. Emails queue. Meetings cannot be joined. Calendars cannot be checked. People do not know whether their messages were sent, whether they were received, or whether they will ever arrive.
The disruption hit businesses and government agencies across several countries, according to reporting on the incident. That is a broad statement, and the specific organisations affected have not been named in the reporting we reviewed, but the breadth of the outage, 74 countries and territories within an hour, makes the breadth of the impact unsurprising. Any organisation that depends on Microsoft 365 for its daily work was affected, and that is a very large number of organisations.
The timing made it worse. The outage began in the afternoon UTC, which meant it hit the European working day in the late afternoon, the American working day in the morning and early afternoon, and the Asian working day in the evening. There was no timezone that escaped it. For workers in the Americas, the outage covered much of the working day. For workers in Europe, it covered the end of the day, when emails are sent and meetings are wrapped up. For workers in Asia and Australia, it began in the evening and continued overnight, which meant they woke to a backlog.
The Copilot dimension added a particular frustration. Microsoft has spent the last two years telling workplaces that Copilot will make them more productive, and it has been rolling AI assistance into every part of Microsoft 365. When Copilot goes down along with the email it is meant to summarise, the promise of AI productivity is undermined in a very visible way. That is not a technical point; it is a credibility point, and it is one reason the outage attracted so much attention. A company that sells reliability as well as intelligence has to deliver both.
What Microsoft Did Right, And What It Owes
It is fair to give Microsoft credit where it is due. The company acknowledged the incident quickly, within minutes of the first reports, and it provided regular updates through its Microsoft 365 Status account and the admin centre. It identified a technical cause within a day, it apologised, and it promised a full post-incident report. The apology came from a senior executive, Charles Lamanna, who said the company was reviewing the incident and working to prevent similar disruptions. The recovery, while not instant, was substantial within 12 hours, and the mail timeout extension was a sensible step to protect queued messages.
It is also fair to say what Microsoft owes. A full post-incident report, with a detailed root cause, a timeline, and specific measures to prevent recurrence, is the minimum. The report should explain exactly which network change exposed the earlier defect, why the health-check system allowed a cascade, and what has changed in the deployment and testing process. It should also address the data question directly: whether any customer data was exposed, lost or delayed beyond recovery, and what the company will do about it.
The broader question is whether this incident, the third major Microsoft 365 disruption of 2026, points to a systemic problem. The earlier outages this year included multi-hour disruptions in January and a code-update failure in August that took several days to fully resolve. Three major outages in one year, with different causes, do not prove a single underlying failure, but they do raise a question about the resilience of the platform. Microsoft’s cloud is one of the largest and most complex systems ever built, and its very scale makes it vulnerable to the kind of cascading failure seen on 12 November. The company’s answer will need to be more than a fix for this specific defect; it will need to be a credible account of how it prevents cascades in general.
What Users And IT Teams Should Do
For individual users, the practical advice is simple. If your Outlook or Teams is still not working, check Microsoft’s official Service health dashboard, and if you are an administrator, check the incident references EX1550543 for Exchange Online and MO1550821 for the wider Microsoft 365 incident in the admin centre, or at the status website Microsoft provides when the admin centre is unavailable. Do not assume the problem is on your side. Restart your Outlook client if you have not already, because a client that was connected during the outage may need to reconnect. And check whether emails sent during the outage were actually delivered, because queued messages may have been delayed.
For IT teams, the outage is a reminder to have a plan for the next one. That plan should include a way to communicate with staff when email itself is down, a way to check service status without relying on the affected platform, and a clear statement to staff about what to do with time-sensitive messages during an outage. It should also include a review of how much the organisation depends on a single cloud provider for email, meetings, documents and AI assistance, and whether that dependence is acceptable. The answer will be different for every organisation, but the question is worth asking after a day when all of those services failed together.
For organisations evaluating Copilot and Microsoft’s AI offerings, the outage is a data point, not a verdict. Copilot was affected because it depends on the same infrastructure as the services it assists, and that dependence is inherent to the product. The question to ask is not whether Copilot can fail, because it can, but whether the value it provides when it works outweighs the risk when it does not. That is a judgement for each organisation, and it should be made with the outage in mind.
For readers who want to understand Microsoft’s AI strategy and its cloud business more broadly, our coverage of Microsoft Azure and AI spending explains the scale of the investment behind the platform, and our guide to what Copilot is explains what the assistant does when it is working. Our earlier report on the Microsoft staff response to the AI push captures the mood inside and around the company on the very day the outage struck.
The Pattern: Cloud Outages Are Infrastructure Stories
The Microsoft 365 outage is the latest in a long series of cloud outages that follow the same pattern. A change is made. A defect is exposed. A health-check system, designed to protect the platform, takes servers offline. Traffic shifts. The survivors are overwhelmed. The cascade spreads. The recovery is gradual. The post-incident report promises lessons learned.
That pattern is worth understanding, because it explains why cloud outages are so hard to prevent. The systems are too large and too interconnected for any one person to understand fully, and the safety mechanisms that are meant to contain failures can themselves become the mechanism by which failures spread. A health check that removes a sick server from service is doing its job, but if too many servers fail their checks at once, the health-check system becomes the cause of the outage rather than the cure.
The industry’s answer to this problem is redundancy, gradual rollouts, and the ability to roll back changes quickly. Microsoft’s response included freezing routine deployments and configuration updates, which is the standard way to stop making a bad situation worse, and deploying an alternative health-check method so that restarted servers could rejoin safely. Those are sensible steps, and they are the steps any large cloud operator would take. The question for the post-incident report is why the defect was not caught before it reached production, and what testing or rollout change would have caught it.
For users, the lesson is simpler: cloud services fail, and they fail in ways that are hard to predict. The right response is not to abandon the cloud, which is impractical for most organisations, but to plan for its failures. That means having backup communication channels, knowing how to check service status, and understanding which of your workflows depend on a single provider. The Microsoft 365 outage was a reminder of that dependence, delivered on a global scale, in the middle of a working day.
What We Could Not Verify
- Microsoft has promised a full post-incident report, but it has not been published yet. The detailed root cause, the exact change that exposed the defect, and the measures to prevent recurrence are therefore unconfirmed, and the account in this article may be revised when the report appears.
- The incident start time is reported slightly differently by different outlets: Microsoft’s Exchange Online incident is reported as beginning at 16:17 UTC by one outlet and the incident is described as opened at 16:20 UTC by another. We have not verified the exact minute, and we use the late afternoon UTC as the timing.
- The figure of 74 countries and territories comes from reporting on Downdetector data. We have not verified the Downdetector figures directly, and the number of countries with reports is not the same as the number of countries with confirmed service disruption.
- The Downdetector figure of roughly 32,000 reports by the evening of 12 November is reported by one outlet. We have not verified it against Downdetector directly.
- The claim that this is the third major Microsoft 365 disruption of 2026, following outages in January and August, comes from reporting on the incident. We have not independently verified the January and August incidents.
- We have found no evidence that AI caused the outage, and Microsoft has said explicitly that it did not. That is a statement about the evidence available now. If the post-incident report identifies a different cause, this article will need to be updated.
- The specific organisations affected, including businesses and government agencies, have not been named in the reporting we reviewed, and we do not name any.
Sources
- Outlook India, “Microsoft Outlook Outage Explained: What Caused It And Is It A Cyberattack?”: https://www.outlookindia.com/amp/story/international/microsoft-outage-cause-cyberattack-server-failure-executive-clarifies
- RansomNews, “Microsoft 365 outage: was it a cyberattack? What we know”: https://ransomnews.com/microsoft-365-outage-cyberattack-november-2026/
- The Hindu, “Microsoft Outage LIVE: Microsoft apologises as global service outage eases”: https://www.thehindu.com/sci-tech/technology/microsoft-outlook-outage-status-microsoft-365-services-live-updates-november-14-2026/article71700692.ece
- BusinessToday, “Microsoft 365 disruption hits 74 countries; Outlook, Teams and Copilot knocked offline”: https://www.businesstoday.in/amp/technology/story/microsoft-365-disruption-hits-74-countries-outlook-teams-and-copilot-knocked-offline-567213-2026-11-13
- NextNews, “Microsoft 365 Outage Update: Cause, Recovery and Outlook Fix”: https://nextnews.com.au/how-to/microsoft-365-outlook-teams-global-outage-november-2026-what-to-do/
- Astra Security, “Microsoft 365 Outage: What Happened and What I’d Do Next”: https://www.getastra.com/blog/security-advisory/microsoft-365-outage/