A business can do everything right at its end and still lose access to a service it depends on. Microsoft 365 provided a reminder on 12 November 2026, when email and several connected services became unavailable to many customers. Over the next few days, technical failures affected US flights, banking services, an NSA internal network and Apple products.
The incidents were not one coordinated attack. Nor is there evidence that they all had the same underlying cause. What makes them worth discussing together is the difficulty of changing complex systems without disrupting the people who rely on them.
Your company website is a smaller system, with the same basic problem. Software needs updates. Some updates can break things. Leaving known vulnerabilities unpatched is not a good alternative.
The outages, without the dramatic headlines
Microsoft, 12–13 November. Exchange Online and other Microsoft 365 services were affected for almost a day. Microsoft's incident review describes a latent defect exposed by changes in traffic and available capacity. The company said it found no cyberattack behind the failure.
US aviation, 13 November. The Federal Aviation Administration imposed a nationwide ground stop after instability in its communications network. Work involving a replacement router at one location was among the factors investigated. The FAA found no indication of a cyberattack and said critical flights and safe landings continued.
Chase, 16 November. Some banking services experienced disruption. The bank attributed the incident to a software problem in an internal hardware system rather than malicious activity.
NSA, 17 November. The agency reported temporary degradation of internal systems during routine maintenance, while saying its mission-critical capabilities remained operational. News reporting connected the problem to software patching.
Apple, 17 November. Customers reported problems with iCloud, Apple Pay and other services. Apple attributed the interruptions to an internal update, not a cyberattack.
These descriptions matter. Saying that an entire organisation was 'down' can suggest far more than the available evidence supports. The useful question is what failed, which users were affected and what the organisation has actually confirmed.
Security alerts were arriving at the same time
The outages were not the only technology news that week. Several genuine security problems also needed attention:
- Attackers abused a vulnerability in Optimizely Web Experimentation on 16 November to insert malicious JavaScript into some customer projects. The affected code could then be delivered to visitors of websites using that service. Optimizely said it had closed the vulnerable path and removed the injected code.
- Java4Shell (CVE-2026-114938) was disclosed on 17 November. It concerns specific Java runtimes and conditions, so administrators need to check the affected versions rather than assume every Java application is exploitable.
- NGINX issued fixes for vulnerabilities including CVE-2026-123007, affecting HTTP/2 processing in vulnerable configurations. CERT-EU advised prompt updates of affected installations.
- On 19 November, PHP released security updates for supported branches: 8.2.36, 8.3.37, 8.4.28 and 8.5.13. An advisory for CVE-2026-129481 explains that a previous fix had been incomplete in PHP 8.3 and later, under particular conditions.
These security advisories should not be presented as the explanation for all five outages. They illustrate a second challenge: administrators have to respond to real vulnerabilities while keeping working services available.
What to ask whoever maintains your website
You do not need to follow every vulnerability bulletin yourself. You do need clear answers to five practical questions.
- Who is responsible for each layer? A hosting company may maintain the server without updating WordPress, plugins, a custom application or third-party integrations.
- Which versions are actually installed? Ask for a short inventory. For a WordPress website, the administration dashboard and hosting panel provide much of the information.
- Has anyone tested a backup? A backup file is useful only if it can be restored. A test restoration is more convincing than a reassuring dashboard icon.
- What happens if an update goes wrong? Where practical, test it on a staging copy. For urgent security fixes, have a verified backup and a plan to restore service quickly.
- Which external scripts still earn their place? Analytics, advertising, chat and testing services can all run code on your pages. Remove integrations you no longer need.
If your website uses PHP, our guide to PHP support dates and the questions to send your host covers the immediate checks.
Two common questions
Would it be safer to stop updating the site? No. That might avoid a short-term compatibility issue, but known security flaws would remain. Controlled updates, checks and recovery plans are a better response.
Does automatic hosting maintenance cover everything? Not necessarily. Ask whether it covers the operating system, PHP or other runtimes, the application itself and its plugins. Those are often separate responsibilities.
We maintain websites that we did not originally build, including update testing and backup checks. If you are unsure who looks after yours, send me its address. I can explain what an external inspection can and cannot tell us.
Sources: Microsoft 365 incident record · FAA ground-stop report · Reuters on the Apple outage · CBS News on the NSA disruption · BleepingComputer on Optimizely · Microsoft's Java4Shell guidance · CERT-EU on NGINX · PHP changelog · PHP advisory for CVE-2026-129481.
