Connect with us

NEWS

The Microsoft Outage That Reached ASIO After the Pact

ASIO and the AFP lost a Microsoft 365 layer, and Canberra activated its June pact as Exchange Online stayed unstable for more than 17 hours.

Published

on

A Microsoft 365 outage that began at 16:17 UTC on November 12 reached Australia’s spy service and federal police, and Canberra activated a June 2026 pact with Microsoft to get priority restoration. Exchange Online was the break. Outlook, Copilot and other Microsoft 365 apps that reach into mail went with it.

Microsoft said it had no sign of a cyber attack. Restored servers kept failing once traffic returned. By 9:24 a.m. UTC on November 13, more than 17 hours after the start, the company still described only early stabilization.

Exchange Online Took 19 Services With It

Microsoft’s service health dashboard timed the Exchange Online incident at 16:17 UTC on November 12. Users could not reach mailboxes through Outlook on the web, desktop Outlook (MAPI), Exchange ActiveSync or REST, and delivery lagged. The first public note from Microsoft 365 Status, at 17:54 UTC, confined the probe to the Americas. Thirteen minutes later the company said users in any region could be hit.

The failure was not 19 separate crashes. It was one mail and calendar layer, felt wherever another product reaches into that layer. Microsoft listed 19 Microsoft 365 services as affected through that dependence, under incident MO1550821, with the Exchange ticket carried as EX1550543.

THE OUTAGE IN FIGURES

  • Start time: 16:17 UTC on November 12, from Microsoft’s service health dashboard.
  • Peak reports: Downdetector logged 138,058 Outlook problems around 1 p.m. ET on November 12.
  • Geographic spread: reports reached 74 countries and territories in about an hour.
  • Mail hold: Microsoft extended email timeout to 48 hours on November 13 to keep delayed messages.

Downdetector later showed those Outlook reports falling to about 34,490 by evening in the United States. Australia peaked above 5,500 Outlook reports at 9:22 a.m. AEST on November 13. The tracker undercounts because most people never file a report, but the shape is clear: the break moved with the working day.

Microsoft told administrators that Teams core messaging and online meetings stayed largely up, while calendar access and meeting scheduling could fail. Direct files in SharePoint Online still opened. Copilot Chat might not load, and Copilot could not read work mail, calendars or files. Defender for Office 365 Safe Links might not check URLs clicked in Outlook and Teams.

SELECTED IMPACTS ON THE MAIL LAYER

ProductWhat failedWhat Microsoft said still worked
Exchange Online / OutlookMailbox access on web, desktop, mobile and REST; delayed deliveryFeatures that do not call Exchange
Microsoft TeamsCalendar and scheduling that depend on ExchangeCore chat and online meetings, largely
SharePoint OnlineFeatures that depend on ExchangeDirect file access
Microsoft 365 CopilotChat load, replies, and work data in mail, calendars and filesNot stated as separately healthy
Defender Safe LinksURL checks inside Outlook and TeamsOther Defender functions not listed as down

Intune enrolment and compliance, Purview labelling and data-loss rules, Planner, To Do, Graph calls for Exchange data, parts of Power BI and Fabric, Microsoft 365 Education tools, admin-center pages, Universal Print and some Windows 365 Cloud PCs that use Purview Customer Key were also named. Sanchit Vir Gogia, chief analyst at Greyhound Research, said firms had priced Microsoft 365 as a licence and never as a single point of failure, and that a rehearsed fallback for incident command and critical mail is the gap that shows up in a night like this.

ASIO, the AFP and a Partial Blackout

Australian Security Intelligence Organisation Director-General Mike Burgess and Australian Federal Police Commissioner Krissy Barrett stood with Home Affairs Minister Tony Burke on November 13 and said both agencies had been hit. Burgess said the spy service operates on multiple levels and had been affected on one level, yes, but not completely. Barrett said the federal police could work around the break on other platforms.

We operate on multiple levels.

Mike Burgess, Director-General, Australian Security Intelligence Organisation, news conference

That is a workaround, not a second inbox. Burke said he had spoken with Jane Livesey, president of Microsoft Australia and New Zealand, and with Lieutenant General Michelle McGuinness, the National Cyber Security Coordinator. Microsoft’s line to him was that this did not look like a cyber event, and that it looked like an update that had gone awry, in some ways similar to the CrowdStrike crash of July 2024.

Charles Lamanna, Microsoft’s executive vice president of Copilot, Agents and Platform, wrote on X that the company regretted the disruption and that there was no indication of malicious or cyber activity. Status updates used a slightly cooler phrase, that Microsoft was not aware of malicious activity causing the incident. Neither line is a finished root-cause finding. Microsoft has promised a post-incident report once the break is mitigated.

Schools, courts and government offices in Australia lost Outlook and Teams. The Australian Medical Association said some hospitals and clinics had clerical trouble, with no reported harm to patient care. The ACT, Tasmania and Western Australia governments said they were hit. Service NSW told people some counters were unavailable. In Japan, Deputy Chief Cabinet Secretary Masanao Ozaki said ministries spent Friday’s working day struggling to send mail and that staff were making physical copies of documents. Singapore’s Government Technology Agency said government mail was down. The Supreme Court of the Philippines warned on X that replies to emails and other transactions could be delayed.

Canberra Signed Its Microsoft Pact in June

Burke said Australia activated the first of the formal cyber agreements he had signed, the one with Microsoft, so Australian interests would get “appropriate prioritisation” in the restore. The paper behind that claim is the June 2026 digital resilience pact with Microsoft, announced by Home Affairs as a landmark deal covering cloud systems, government platforms and AI.

Microsoft’s own newsroom dated the Canberra signing June 9, 2026, with Burke and Lisa Monaco, president of Microsoft Global Affairs, as the signatories. The published memorandum is a non-binding framework for joint work, not a contract that names a restore SLA. Five months later its first public use was a place in the queue after Exchange failed.

WHAT THE JUNE PACT NAMED AS FIRST JOBS

  • Subsea cables: physical and cyber security of the links that carry Australian traffic.
  • Hyperscale cloud: resilience of the large cloud platforms government now runs on.
  • Secure AI: rules for how government adopts AI without opening new holes.
  • Small firms and utilities: help for businesses and critical infrastructure that sit outside the federal estate.
  • Emerging threats: a standing dialogue, with working groups expected to meet at least once a year.

The same Microsoft notice tied the pact to a record A$25 billion Australia investment announced in April 2026 during a Sydney visit by Satya Nadella, plus a plan to train three million Australians in AI skills and to extend the Microsoft-ASD Cyber-Shield to more agencies. Jane Livesey called resilience a growth strategy. On November 13 the growth strategy was a phone call to get Australian mail moved up the list.

Burke’s June line was that the country cannot stop all cyber attacks, and that deals like this make it more resilient. The November incident was not, on Microsoft’s present evidence, an attack. The pact still had to be switched on because so much official work sat on one mail system.

8.5 Million Devices Did Not Change the Stack

On July 19, 2024, a faulty CrowdStrike Falcon sensor update crashed Windows machines worldwide. Microsoft estimated about 8.5 million Windows devices, less than 1 percent of all Windows machines. Airports, 911 desks, hospitals and banks stalled. Thousands of flights were cancelled. Delta’s crew-tracking software became a second failure that outlasted the first. Burke put that night next to this one because both, in his briefing, looked like an update that had gone wrong.

They are not the same break. CrowdStrike knocked the operating system into a restart loop. This incident left PCs running and took the shared mail layer those PCs talk to. Planes still flew at IndiGo and Air India. The bill arrived in inboxes, courts, school platforms and the security agencies that had kept a Microsoft layer in production.

TWO OUTAGES, TWO FAILURE POINTS

CrowdStrike, July 19, 2024Exchange Online, November 12, 2026
What brokeWindows kernel crash after a Falcon content updateExchange Online mailbox access and every app that calls it
Microsoft’s countAbout 8.5 million devices, less than 1% of WindowsNo user total published; 19 Microsoft 365 services listed
AviationGrounded check-in and, at Delta, crew trackingIndiGo and Air India kept flying; support mail lagged
National securityEmergency and hospital systems among the wreckageASIO and AFP hit on one layer, with other platforms still up
Vendor storyCrowdStrike logic error, later a 20-versus-21 field mismatchRoot cause unnamed; health checks and a secure-connection process blamed in flight

The 2024 lesson that circulated in government was concentration risk. Australia’s 2026 answer was a closer Microsoft partnership, more local cloud spend and a cyber pact. Diversity of vendors was not the product. Priority restoration was.

Why Restored Servers Keep Going Dark

Microsoft’s first technical story was elevated load. It then said the extra load was not new user traffic. A large number of servers had dropped out after failing automated health checks, which shoved work onto the machines still up. A process that sets up secure connections was becoming overwhelmed, so those servers stopped handling requests. Engineers paused deployments, shifted traffic, restarted machines and tested a new health-check method.

The ugly loop arrived overnight. Servers brought back to service went unresponsive again once they took traffic. Microsoft began limiting which traffic could reach damaged kit in some locations and rebuilt a subset of servers on an earlier configuration. At 3:13 a.m. UTC on November 13 it said the changes meant to improve serviceability were not taking as expected. Virtual machines spun up to add capacity got stuck in a bad state, so the new settings would not sync.

HOW MICROSOFT’S STORY MOVED

  1. November 12, 16:17 UTC: Exchange Online incident opens on the service health dashboard.
  2. November 12, 17:54 UTC: Microsoft 365 Status says it is investigating mailbox access in the Americas under EX1550543.
  3. November 12, 18:07 UTC: the company says users in any region may be affected.
  4. November 12, 19:31 UTC: elevated load is blamed on servers that became unexpectedly unavailable, not on a traffic spike.
  5. November 12, 19:50 UTC: a configuration test targets servers that fail automated health checks.
  6. November 12, 21:00 UTC: no recent change is tied to the start; a health-processing component is suspected.
  7. November 12, 22:09 UTC: actions so far have not provided the anticipated relief; a core process is overwhelming capacity.
  8. November 13, 00:06 UTC: servers returned to service become unresponsive again under traffic.
  9. November 13, 03:13 UTC: serviceability changes are not taking as expected.
  10. November 13, 09:24 UTC: early stabilization is reported and email timeout is stretched to 48 hours.

That 00:06 UTC note is the mechanism in one sentence. Capacity that looks healthy in isolation does not stay healthy once the world reconnects. Readers who opened the Microsoft 365 status page saw only the latest paragraph; the RSS feed kept the trail of reversals.

Login systems that send one-time codes to Outlook were stranded with the mail, which is how an email outage becomes an identity outage. Japan’s paper copies and Australia’s “talk to people” jokes were the same fallback wearing different clothes.

After the Planes, the Inboxes

IndiGo said the outage hit its email support desk, so it could not take customer queries that way, while other support channels stayed up and flights were not affected. Akasa Air told passengers to expect delays on email support. Air India said it had moved mail onto other routes, and that flights and customer channels were not hit. South Africa’s CemAir said it was seeing system disruptions from the global outage, a broader claim than the Indian carriers made.

That split is the second-order map. Check-in and crew tools were not the first things to die, which is why this night did not look like July 2024 at the airport. The work that lives in Outlook did die: support queues, court mail, school messages, city-hall inboxes. Several U.S. city governments told residents to phone. Workhorse Group postponed its quarterly earnings call. In Hong Kong, staff at Morgan Stanley and Goldman Sachs lost Teams and mail; one banker said some messages dribbled in while a backlog of millions could be stuck.

OpenAI said it was investigating a Microsoft connector problem in which new connections may be broken and tools do not work as expected. Anthropic said an upstream provider issue was hitting its Microsoft 365 connector, with mail, SharePoint, Teams chat and calendar requests failing. ChatGPT and Claude as whole products were not the outage. The pipe each one uses to read a customer’s work tenant was.

Copilot, ChatGPT and Claude Lost the Same Feed

Copilot is Microsoft’s own assistant on that same mail layer, so it went dark for the same reason Claude and ChatGPT did. Safe Links, the check that is supposed to run when someone clicks a URL in Outlook or Teams, was also on the list, which means a security control can fail open at the same moment the inbox does. Gogia’s point about writing down whether each control fails open or closed is the practical residue of the night, and it does not need a second office suite to be true.

At 9:24 a.m. UTC on November 13, Microsoft said it was seeing early stabilization and had stretched email timeout to 48 hours so delayed mail would not be dropped. It still owed customers the post-incident report that names the root cause and the repairs behind it. Canberra had already spent its pact on a place in that line.

Harry is the editor of NEWS ANALYSIS. He writes across the publication's ten desks, with most of his time going to the stories where a number, a filing or a study decides the argument. His working rule is simple: read the source document before writing about it, and tell the reader plainly which parts are established and which are somebody's claim. He is responsible for the standards set out on this site's Editorial Standards and Fact Checking pages, and for correcting the record openly when the publication gets something wrong.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending