Nov 19, 2026 · 8 min read
Microsoft 365 Outage Phishing Used SendGrid Tracking Links
On November 13, while Exchange Online was still climbing out of a global outage, a phishing campaign told Microsoft 365 users that their held messages would be deleted within 24 hours. Every button ran through recipient specific click tracking links, including SendGrid hosts, before landing on a site that has nothing to do with Microsoft.
The outage was real, so the lure did not need much invention. Exchange Online incident EX1550543 started on the afternoon of November 12, 2026, and Microsoft's preliminary post incident report puts its length at 23 hours and 50 minutes, according to Tony Redmond's analysis at Office 365 for IT Pros. Delayed mail was still trickling into inboxes the next morning. That same day, Microsoft Threat Intelligence saw a "coordinated, high-volume phishing campaign" that dressed itself up as a Microsoft 365 service health notice about exactly that missing mail.
Key Takeaways
- Microsoft Threat Intelligence says it observed the outage themed phishing emails on November 13, 2026, one day after Exchange Online incident EX1550543 began.
- The lures claimed messages had been held during the outage and would be deleted within 24 hours, and pushed a "RESTORE MESSAGES" button.
- Three sender domains carried the campaign, moments-notice[.]com, planetmaison[.]com and newbooker[.]com, and the clicks passed through tracking hosts that include ct.sendgrid.net subdomains before reaching reallifeaccounting[.]com.
- Microsoft assesses the activity as likely credential phishing but says its evidence does not confirm credential capture, and it gives no message count.
- Microsoft's own fix for delayed mail went the other way: it extended its email timeout window to 48 hours, per BleepingComputer, so real queued mail was being kept longer, not deleted.
What Did the Fake Outage Emails Say?
The emails claimed a handful of your messages were stuck and gave you one day to save them. According to the Microsoft Security Blog post, one sample referenced "a purported Exchange Online incident on November 12 and claimed that four messages were held after the outage." The details Microsoft published:
- Subject lines: "Outage" plus a random six character token, for example
Outage – aF7vG6. Microsoft also lists two other randomized formats:Strengthen – Security <random>andRestore — 4 messages <random>. - Display names: a three letter, organization style prefix followed by "Outage Notification" or "Service Health," such as "MSF Outage Notification" or "AMZ Service Health."
- Body: a Microsoft 365 service health look, with mentions of Exchange Online, Outlook, Teams, SharePoint and Copilot, plus buttons labelled "RESTORE MESSAGES" or "TAKE ACTION."
- Filter bait: a fake email conversation hidden from the reader, which Microsoft says "appears to be an attempt at mail filtering evasion."
Microsoft found no QR codes, no malware and no payload attachments. It was a link campaign from start to finish.
Why Did the Lure Work So Well During This Outage?
It described something users were actually experiencing. During the outage, BleepingComputer reported that Microsoft's advisory MO1550821 warned customers they might be unable to reach Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot and Microsoft Defender XDR. Exchange users told the site on November 13 that delayed emails had begun "trickling into their inboxes." A notice saying four of your messages were held fit that picture perfectly.
The deadline is the tell. Microsoft responded to the backlog by extending its email timeout window to 48 hours "to help preserve delayed messages," BleepingComputer reported. The phish claimed the opposite: a 24 hour countdown to deletion. Half the real window, and a threat Microsoft never made. Microsoft's own guidance in the blog post is blunt: legitimate service health communications "do not direct them to non-Microsoft domains to recover held mail."
Microsoft also confirmed the outage itself was not an attack. A company statement to BleepingComputer called it "a technical issue not related to cyber or AI / agents." The attackers simply moved fast enough to ride it. We flagged this risk during the June 2026 Exchange Online outage; this time Microsoft observed the lures while the incident was still running.
How Did SendGrid Click Tracking Hide the Destination?
The buttons did not point at the phishing site; they pointed at click tracking hosts that forward you there. Microsoft says the campaign "used email-delivery and link-tracking infrastructure to generate recipient-specific URLs," and its indicator list names three u…[.]ct[.]sendgrid[.]net hosts alongside url[.]newbooker[.]com and url7489[.]newbooker[.]com. Its hunting query matches the paths u113727626.ct.sendgrid.net/ls/click and url.newbooker.com/ls/click.
That is standard email service plumbing. SendGrid's tracking documentation says "Click tracking rewrites all the links and URLs in your emails," and that the links "then point to either Twilio SendGrid's servers or the domain you branded your link." Its link branding guide explains that branding "replaces sendgrid.net" with the sender's own domain, which matches the url[.]newbooker[.]com hosts sitting next to the sendgrid.net ones. Microsoft does not say whether newbooker[.]com is attacker owned or a compromised account, so we don't either.
Here is the problem for the reader. Hover over the button and your mail client shows a tracking host, not reallifeaccounting[.]com, outagefixed[.]sbs or outagerestored[.]sbs, the downstream sites Microsoft observed. Filters face a related problem: the visible host belongs to a delivery platform that thousands of real senders use. Microsoft warns defenders to "avoid broadly blocking shared email-delivery infrastructure based only on this campaign," because doing so would break legitimate mail.
The links were also per recipient. Microsoft says a "recipient-specific tracking URL records or routes the click," so opening the link, even from a sandbox, can tell the sender that your copy was clicked. The indicator list adds an individualized px[.]png image on 162[.]120[.]17[.]69:8080. That is an open tracking pixel, the same trick we covered in how attackers use tracking pixels to find live inboxes.
Why Email Users Should Care
The redirect layer that hid this phish is the same one in most of the marketing and transactional mail you get. Our breakdown of how SendGrid tracks your email shows what a normal tracked link looks like: a SendGrid or branded host up front, the real destination resolved only on SendGrid's side. Most people click links built that way every week, so a phishing link built the same way looks routine. Hovering, the classic advice, tells you little when legitimate links hover exactly the same way.
This campaign targeted work mailboxes, but the lesson carries to Gmail. Any big, public service failure, at Google or Microsoft or your bank, is a ready made pretext, and a tracked link means the sender learns who clicked before you have decided anything. We have seen the same move with phishing sent through legitimate platforms like DocuSign and QuickBooks: trusted infrastructure carrying an untrusted message.
How Can You Check a Suspicious Outage Email?
Skip the email and check Microsoft's status pages directly. Microsoft's service health documentation says admins can see incidents under Health, then Service health, in the Microsoft 365 admin center, and that anyone locked out can use the public status page at status.cloud.microsoft.
- Read the sender domain, not the display name. "MSF Outage Notification" is just text. A notice sent from moments-notice[.]com or planetmaison[.]com is not Microsoft.
- Treat any deletion deadline as a red flag. Microsoft kept delayed mail longer during this outage. It did not threaten to delete it.
- Don't click to find out where a tracked link goes. The URL is unique to you, so the click itself is a signal to the sender. Ask your IT team, or report the message through Outlook's reporting button as Microsoft recommends.
- Use phishing resistant sign in. Microsoft recommends prioritizing phishing resistant methods such as passkeys or security keys, which are built so they cannot be replayed on a fake login page.
What Should Microsoft 365 Admins Do Now?
Hunt for the messages, find who clicked, and reset anyone who may have typed a password. Microsoft's post says Defender for Office 365 already blocks and quarantines the messages, but tenants without it, or with copies delivered before detection, should act:
- Run the hunting queries, then widen them. Microsoft's first query lists moments-notice.com and newbooker.com but not planetmaison.com, which its own indicator table includes, and its subject regex covers only the "Outage" format. Add planetmaison.com and the "Strengthen – Security" and "Restore — 4 messages" subjects.
- Trace the redirects. In the UrlClickEvents table, the UrlChain column "includes URLs present in the redirection chain," which is how you see past the tracking host to the destination.
- Reset and revoke. For anyone who may have entered credentials, Microsoft says to reset the password, "revoke active sessions and refresh tokens," verify MFA registrations, and review OAuth consent, inbox and forwarding rules and delegate permissions.
- Block the destination, not the platform. Block or monitor reallifeaccounting[.]com and its /themelist0.html path. Judge the sendgrid.net hosts with sender and content context.
- Turn on cleanup. Microsoft recommends Safe Links, Safe Attachments and zero hour auto purge, which pulls back phishing that landed before new threat intelligence arrived.
Microsoft maps the activity to MITRE ATT&CK T1566.002, Spearphishing Link, along with T1036 Masquerading and T1204.001 Malicious Link.
What We Could Not Verify
We did not receive one of these emails, so the lure details come from the samples in Microsoft's write up. Microsoft gives no message count, has not named a threat actor, and says its evidence "did not include the final page source, credential fields, credential-submission telemetry, or post-compromise activity." Microsoft does not say whether the lures cited a real incident number. The 23 hour 50 minute duration comes from Microsoft's preliminary post incident report as quoted by Office 365 for IT Pros; we have not seen the report ourselves.