decrypted · 16 november 2026 · vulnerabilities and patching · ai and llm security · digital sovereignty

A Microsoft outage showed how one network change can stop your email

For more than a day this week, a large part of the working world could not rely on its email. From Thursday evening, UK time, Exchange Online, the Microsoft-run mail service behind Outlook, stopped working for customers worldwide, and Teams, Copilot and Defender for Office 365 went with it. Microsoft says it was not an attack. That is the interesting part: nobody needed to break in for a global outage to happen.

What Microsoft says happened

Microsoft's account, as it emerged through status updates, is a chain. An earlier network capacity upgrade left a fault that surfaced after a later network change. Servers began failing their automated health checks and dropped out. The remaining servers took the extra load, and a core process on them became overwhelmed too. Microsoft said it was a technical issue not related to cyber or AI agents, and that it found no sign of malicious activity.

Picture a motorway where one lane closes. Traffic floods the others, which jam in turn, and soon the whole road is stationary. Restarts helped, but The Register reported that some recovered servers could become unresponsive again once traffic returned, which is the jam re-forming as cars rejoin.

The controls sat on the same road

Administrators also lost access to Intune and the Microsoft 365 Apps Admin Center, the tools used to manage devices and software, according to The Register. The controls for responding shared the fate of the thing they were meant to fix. Microsoft extended its email timeout window to 48 hours, which gives queued mail longer to arrive, but also means a backlog that drains slowly.

Dame Chi Onwurah, chair of the Commons Science and Technology Committee, said the outage "only serves to highlight once again the lack of resilience in the UK's digital infrastructure". That is a fair reading: many UK organisations have quietly made one supplier the single route for staff, customers and administrators alike.

The Secure by Design lesson

Secure by Design is not only about keeping attackers out. It asks what happens when a component fails, and whether the failure stays small. Three decisions on the customer side would have limited this one:

None of this is expensive. Most of it is a rehearsal and a one-page list. The costly version is learning it live, on a Friday, with clients ringing.

Microsoft has promised a full post-incident report once its review is complete. Boards should ask for it, and read it for the root cause rather than the apology.

Also this week

AI-driven intrusions. The FBI has issued a flash alert on ARTEX, an open-source framework that connects LLMs, the large language models behind chatbots, to hacking tools so software agents can run an intrusion from reconnaissance to data theft. The FBI says some intrusions moved from first targeting to privileged access in under three hours, and that it has been used since at least August 2026. The author withdrew the code on 8 October, but forks and a commercial version persist. Reporting links it to October's compromise of ConnectWise's cloud backup service, where attackers used compromised application identities to reach customers' Microsoft 365 tenants; Microsoft attributed that to a China-state-aligned group, and ARTEX was used to try more than 2,250 downstream tenants. The tool hunts for forgotten partner portals and third-party links, so an accurate inventory of what you expose is now a security control. Rate-limiting sensitive APIs and watching for bursts of automated requests, as the FBI advises, is cheap by comparison.

Windows zero-day. A researcher known as Nightmare Eclipse has published SignDrop, an exploit for an unpatched Windows flaw that lets a local user become SYSTEM, the highest privilege on the machine, by abusing the Common Log File System and tricking Windows Defender into creating a file with that power. The researcher claims it can switch off EDR, software that watches laptops and servers, and estimates around 80% of Windows installations are exposed, having tested recent Windows 11 and Windows Server 2025 builds. Microsoft says it is investigating, that Defender detects and blocks exploitation, and that it has not seen it used in the wild. The reporting I saw gives no CVE number and lists no patch or workaround. It needs local access first, so it matters most after a phishing foothold. That is worth a conversation with whoever runs your endpoint tooling.

Sources

If you want help working out where one supplier is a single point of failure in your organisation, get in touch.

More like this

Get the next post by email: subscribe to Decrypted. Double opt-in, unsubscribe any time, or take the RSS feed.

Prefer to listen? Decrypted on Apple Podcasts, or paste the podcast feed into any app.

Play here Loads Apple's player when you press it.