Microsoft's 24-hour outage took Copilot down with Outlook. Its apology says AI had nothing to do with it.
Outlook, Teams and Copilot went dark in more than 70 countries. Microsoft's explanation changed between Thursday and Friday, and the company took care to rule out AI agents.
Microsoft apologised on Friday, November 13, for an outage that BBC News says lasted almost 24 hours and hit more than 70 countries. Outlook, Teams and Copilot were all unreachable for many users. The line in the apology that caught my eye wasn’t the “deeply sorry”. It was the next sentence: “This was a technical issue not related to cyber or AI / agents.”
My read: companies don’t usually rule things out unprompted, so Microsoft evidently expected the question.
How the explanation moved
On Thursday, according to SCMP, Microsoft blamed “elevated load” on its infrastructure. Parliament News, summarising reports from The Times and the BBC, says the company described a large number of servers becoming unavailable, which pushed traffic onto the machines still running. It also said a first fix was “not taking as expected”. By Friday the account had a different shape.
| When | What Microsoft said | Reported by |
|---|---|---|
| Thursday | “Elevated load” on its infrastructure | SCMP |
| Thursday | Fix “not taking as expected” | BBC News |
| Friday | Problem introduced “during an earlier network capacity upgrade that surfaced following a recent network change” | Windows Central, quoting Microsoft’s post on X |
| Friday | Services restored gradually, “validating stability at each stage” | Independent, quoting Microsoft |
Those aren’t contradictory. A network change can cause servers to drop out, and dropped servers cause load on the rest. But it does mean the first public explanation described a symptom, and the cause arrived a day later. Microsoft has promised a full post-incident report once its review is complete, and until that exists the root cause is the company’s own short summary.
The numbers
BBC News puts the affected user base at more than 430 million people who use the Microsoft 365 apps worldwide, citing Microsoft’s website. SCMP, citing the outage tracker Downdetector, counted 74 countries and territories. Those are different measures, and neither tells you how many people actually lost access, which nobody has published.
The reported damage was mostly everyday disruption. BBC News reported UK GP surgeries asking patients to phone instead of using web forms, police forces telling people to call 101 for non-urgent reports, and pharmacies struggling to reach GP practices. BBC News also reported courts, schools and council offices without email. The UK cyber minister told the BBC that 999 calls and frontline NHS 111 services weren’t affected, and the NHS Alliance chief executive said clinical systems appeared to be working.
Why Copilot is in the story
Copilot sat in the list of dead services next to Outlook and Teams, which is the part that matters for this site. BBC News lists Copilot among the services that were inaccessible alongside Teams and Outlook. My read: an assistant that works on your mail and meetings inherits the same single point of failure as the mail and meetings. None of the coverage I read says Copilot had a separate fallback, though that is an absence in the reporting, not proof.
I covered a related argument in an earlier post about AI agent incidents, where Microsoft’s Brad Smith asked for a brake pedal on AI systems. This outage had no AI cause, Microsoft says. My read: it still shows how much new agent software leans on a handful of older services staying up.
My read
My read: Microsoft was right to say what it did not cause, and it’s useful that a company is being that explicit. But “not AI” is a claim I can’t check until the post-incident report lands. I’d treat the network-upgrade explanation as likely and unconfirmed.
The bigger point is the one cyber-security adviser Jake Moore made to the BBC, that heavy reliance on a few providers means one failure spreads far. UK Commons committee chair Chi Onwurah called it a “lack of resilience” in the country’s digital infrastructure. Both quotes are about email and diaries. Now add agents that book, reply and pay on your behalf, and an outage stops being an inconvenience.
What to watch: the post-incident report, and whether it says anything about Copilot’s dependency on Exchange.
Sources
- BBC News: Microsoft ‘deeply sorry’ for global outage after almost 24 hours of disruption
- BBC News: GPs, police and pharmacies among UK services affected by Microsoft outage
- BBC News live coverage: Microsoft outage updates
- SCMP: Microsoft ‘deeply sorry’ for global outage, says issue not caused by cyber or AI agents
- Windows Central: Microsoft apologizes for major outage, insists it wasn’t AI
- The Independent: Microsoft ‘deeply sorry’ for disruption caused by major global outage
- Parliament News: Microsoft outage, GPs, pharmacies, police and schools affected