Was the Microsoft 365 Outage a Cyberattack or AI? What Actually Happened
When Outlook, Teams and Copilot went down worldwide on 12 November, people online quickly blamed hackers, AI agents, or AI-written code. Microsoft says it was none of those. The real attack that week came afterwards, in people's inboxes.
What happened
- 12 November, 16:20 UTC. Exchange Online started failing for Microsoft 365 customers around the world. Microsoft tracked it as incident EX1550543, with a second incident, MO1550821, for the knock-on effects on other services.
- What went down. Outlook and Exchange Online mail, Teams, SharePoint and Copilot. Copilot Chat failed in part because it couldn't reach Exchange Online, where Copilot's compliance records are stored.
- 13 November, 16:10 UTC. Microsoft marked the incident resolved. Some tenants had mail back by about 10:00 UTC that morning.
- 16 November. Microsoft's preliminary post-incident report put the outage at 23 hours 50 minutes. A final report is still to come.
What Microsoft says caused it
According to the preliminary report, as summarised by Tony Redmond at Office 365 for IT Pros, three things went wrong at the same time:
- A latent Windows defect. A Microsoft 365-specific Windows change, deployed globally from July, could cause lock contention under certain loads. That used up CPU and memory on the networking gateway servers.
- Extra traffic. About 15 hours earlier, Teams had failed over to the same primary gateway that Exchange Online uses.
- Less capacity. About 14% of gateway servers were already offline because of an unrelated problem.
A separate fault let the failure cross the regional boundaries that normally keep problems local, which is why the outage was global. Recovery was slow because restored servers were overwhelmed and failed again. Microsoft couldn't simply roll back, because the faulty change had been live for more than three months. In the end it restored the previous known-good Windows code and brought servers back in controlled groups.
Was it a cyberattack?
No, according to Microsoft. Charles Lamanna, the Microsoft executive vice president for Copilot, agents and platform, said there was "no indication this is the result of malicious or cyber activity." Nothing published since points the other way: no group has claimed it, and the fix was a code change, not removing an intruder.
Was it caused by AI or AI agents?
Also no, according to Microsoft. As the outage dragged into a second day, posts on Reddit and elsewhere claimed it came from Microsoft relying on AI in its workflows or cutting experienced staff. Microsoft then said, according to KUOW's reporting, that the outage was not caused by AI or a cyberattack. No one has shown evidence that AI-written code or an AI agent triggered it. Copilot was a victim of the outage, not its cause.
The theory spread because it sounds plausible. AI agents are now part of real attacks, which we cover in Can AI agents hack systems and cause outages?. But this outage has a documented cause, and it is an ordinary one: an old bug, extra load and too little spare capacity.
The part that was an attack: outage-themed phishing
While Microsoft was still restoring service, Microsoft Threat Intelligence spotted phishing emails that used the real outage as bait. According to Packetlabs' write-up of Microsoft's report:
- They looked like Microsoft 365 service notices and quoted the real incident number, EX1550543.
- Subject lines started with "Outage -" followed by a six-character code. Sender names mixed the target organisation's name with "Outage Notification" or "Service Health".
- The message said emails held during the outage would be deleted within 24 hours unless you clicked a link to restore them.
- Links went through SendGrid tracking to reallifeaccounting[.]com. The mail came from moments-notice[.]com, planetmaison[.]com and newbooker[.]com.
Microsoft said the campaign was consistent with credential phishing but had not confirmed what was collected. The pattern is familiar: a real event, a real incident number and a deadline. No real Microsoft notice asks you to click a link to stop your mail being deleted. We cover outage scams in more detail, with a checklist, in Outage scams: what to do when Microsoft 365 or AWS goes down.
If you clicked one of those emails
- Change your Microsoft 365 password from a device you trust, and sign out of all sessions.
- Check your sign-in methods for a phone number or authenticator app you don't recognise.
- Look for new inbox rules that forward or delete mail. Attackers add them to hide replies.
- Tell your IT team, so they can check sign-in logs and block the sending domains.
More steps are in I clicked a phishing link.
Where PhishClean fits, and where it doesn't: nothing on your computer could have prevented this outage. The browser extension checks the page a phishing link opens. It warns, for free, when a sign-in form would send your password to a different site, and during the trial or on Pro it also flags brand sign-in pages on the wrong domain. PhishClean Desktop works on your Windows PC after you leave the page. If a follow-up scam gets you to paste a command, install a remote-access app or run a "repair tool", it steps in.
Sources
- Office 365 for IT Pros, 16 November 2026: timeline, incident numbers, preliminary post-incident report, Lamanna's statement.
- KUOW, 13 November 2026: Microsoft's statements on cyberattack and AI, and the online rumours.
- Packetlabs, 16 November 2026: the outage-themed phishing campaign and its indicators.
Try PhishClean Desktop free for 7 days
Every protection on for 7 days on Windows 10 and 11, no card needed. Your browsing, files and clipboard stay on your PC.
Download PhishClean DesktopLast updated: