top of page
Search

Tech Eggs All In One Basket: A Microsoft And WordPress Lesson Learnt

Nov 13
8 min read

Microsoft's outage began on Thursday 12 November 2026.


At its widest it reached Exchange Online, Outlook, Teams, SharePoint and Copilot. Around 140,000 people in 74 countries reported problems to Downdetector, and airlines and government departments were caught up in it. By the sixteenth hour the fault had narrowed to Exchange, which is another way of saying email.


The cause, on Microsoft's account, was settings applied to servers after an update. Not an attack. A change was made, the change was wrong, and a very large number of businesses found out what their working day runs on.


That is a serious failure by a company that sells reliability. Afterall, the name Microsoft 365 implies it, right?


It is also the least interesting fact of the week.


Cannot connect to the server?
Cannot connect to the server?

What the firms did


First to feel the disruption was Australia.


Meetings moved. Deadlines were marked as 'at risk'.


Clients were told the system was unavailable, in the tone people use for weather.


Some fee-earners sent work through personal email accounts, which is a decision about client confidentiality taken in ninety seconds by someone under pressure. Others stopped and waited for instruction that never came. Very few firms switched to a documented alternative, because very few had written one down.


A firm that cannot send a letter, an invoice or a completion statement for a day has an operational problem. A firm that discovers this for the first time on the day it happens has been trading on an assumption it never tested, while selling risk management to clients as a professional service.


The outage was Microsoft's responsibility. The exposure belonged to the firm, and it was taken knowingly in the way most unmanaged risk gets taken: nobody ever asked.


The excuses, in the order they are offered


"Nobody could have predicted this." Nobody predicted this particular Thursday or Friday. Everyone knew some version was coming, because there has been an outage in each of the last several years and the ones that reach the front pages are a fraction of the total. Forecasting a specific failure is difficult. Allowing for the category is a decision about how you spend an afternoon.


"It's Microsoft's problem." The excuse I find least impressive in a professional firm. Liability can move to a supplier. Consequence cannot. The client whose completion is today does not care whose status page is red, and the duty to act in that client's best interests did not suspend itself while servers were repaired. A supplier's exposure to you is not a plan for your exposure to the client.


"Everyone uses it." True and beside the point. Concentration of the kind Microsoft, Google and Amazon now represent is a recent arrangement in the history of office technology. That your competitors made the same choice means you will all be down at once, which is worse for your clients than failing alone, and worse for you if a regulator later asks what you considered.


"We're too small for this." The argument usually runs the other way. A four-partner firm has one week of buffer and no slack in the payroll, while a 400-person firm can absorb a lost afternoon. Being small makes the question more urgent, and small firms mostly cite it as a reason to defer.


"Duplication costs money." Some of it does. Most of what protects a firm here costs an afternoon: the domain held in the firm's own name, an export opened on a laptop that has never seen the platform, one page describing who calls whom. What firms mean when they say it costs money is that the work is dull and the reward is a day that never arrives.


"The portal dictates our formats." It dictates formats. It does not dictate dependence. You can receive a Word document and still be able to serve that client on a day when the licence check fails.


An unmaintained website is also a choice


The WordPress side of this is where the neglect is easiest to see, because a website is the one system a firm can abandon for years without noticing.


WordPress runs about 39% of all websites, and 57.9% of those whose content system can be identified. That is a genuine achievement for free software and it is also a map. An attacker who finds a flaw does not have to select a victim. The victim is whoever has not updated.


In July the flaws arrived as a pair. In technical blurb, CVE-2026-60137 was a SQL injection in WordPress core which alone required an authenticated user. CVE-2026-63030 was a logic error in the batch REST API that allowed a request to bypass authentication. Chained, they gave an unauthenticated attacker code execution on a default installation. Versions 6.9.0 to 6.9.4 and 7.0.0 to 7.0.1 were affected, and the fix arrived on 17 July in 7.0.2 with forced automatic updates.


None of that means anything to me. It's just technobabble.


Within two days there were more than two dozen public exploits. Honeypots run by watchTowr logged tens of thousands of attempts and over a hundred backdoor administrator accounts created by different attackers. The researchers found the flaws using an AI model in roughly ten hours, which tells you what the next one will cost to find.


Almost no small firm decided to run an unpatched site. A site got built, the plugins made sense at the time, the person who built it moved on, the admin login went into a document nobody opens, and updates happened when somebody remembered. Nobody owns the website. The asset carries the firm's name and collects enquiries from the public, and no named person is answerable for it.


The awkward question is what sits behind the enquiry form. If names, contact details or matter descriptions pass through a site nobody has updated, then a compromise becomes a personal data incident, with an assessment to make inside seventy-two hours about whether it must be reported.


Most firms in that position would not know where to begin looking.


What the client was never told


I have read a fair number of engagement letters. I have not seen one that says the firm's ability to act depends on a vendor whose outage it cannot control and has never tested an alternative to.


That omission is the part I would press on. A retainer sets out the scope of work, the fee basis, who is responsible for what. It rarely sets out that the firm's continuity rests on a single supplier, because stating it would invite the obvious question. Confidentiality obligations and data protection duties attach to the information, not to the computer it is stored on, so the duty to keep client material secure does not subcontract when the storage does.


The quality standards point the same way. Take Lexcel, for example. Continuity arrangements are part of what an assessor can ask to see, and a policy that has never been run is a poor answer.


Professional indemnity proposal forms ask about backup and business continuity every year, and the answers are given quickly, by people who have not checked.


None of this requires an outage to fix. It requires someone to stop treating the question as administrative.


Why Fraxinus was not in it


Fraxinus has escaped this week, as it escaped the last one, and the reasons are worth separating.


The first is luck.


The site is served over a different company's network and the failure was inside Microsoft's mail and collaboration estate. Sitting on somebody else's platform is not a defence in itself, and I am not going to claim otherwise.


The second is the arrangement.


The domain is registered to the firm. The posts exist as files that open without the platform. The address a client writes to is not tied to a single vendor's sign-in. None of that was a project. It is the ordinary way the work is set up, which is the point: contingency that has to be improvised is a document, and contingency that is simply how things are done is a capability.


The standard I would apply to any firm is how quickly it could leave a supplier, and whether anyone has written that number down.


What to do on Monday


  • Write one sentence per critical system. If email is unavailable for a day, we do this. If the site is down, clients hear it this way. Where the sentence ends in "wait", that is your exposure, and it is better found by you than by an assessor or a claimant.

  • Put the domain and the DNS records in the firm's own name, held by the firm rather than an agency or a former employee.

  • Export the content and open the export on a machine that has never seen the platform. A backup nobody has opened is a belief.

  • Separate the vendors where it is cheap to do so. Email, website, document storage and payments failing on the same afternoon is a worse day than one of them failing.

  • Find out who owns the website. If the answer is nobody, that is the finding, and it should be written down and given to a named person.

  • Time the workaround on a normal Tuesday, with the person who would actually be doing it, and write the number down.

  • Keep the record. Continuity is normally answered with a policy, and a policy describes intent. The evidence is a dated note of a test, including the parts that did not work.


My view


The crowd is a distribution of risk that feels like safety because everyone is carrying the same amount. It has never been a defence, and no regulator has accepted it as one.


Microsoft's email is good, WordPress can run a serious website, and the firms using them are not reckless in the ordinary sense. They are worse than reckless in one specific way: they have taken a decision that belongs to the client, about how much disruption the client's matter can absorb, and they have taken it silently, on the client's behalf, without asking. A solicitor who would not move a completion date without instructions has moved the whole firm's ability to act onto one supplier's release schedule, and never mentioned it.


Contingency is a duty of the person who signs the retainer. It is discharged before the outage, in an afternoon nobody will thank you for, and it is either already done when the status page turns red or it is not available at all. A fallback is something you reach for, so someone has to remember it exists, find it, and have the authority to use it while everyone else is explaining the outage. A default happens without a decision. That difference in status is why one gets tested and the other gets written down and filed.


Microsoft will publish its root cause analysis. Most firms will read the summary, agree it was a settings change, and change nothing, because next week something else will be urgent. The ones that come through the next one will have answered the question before this week, and will be slightly embarrassed by how little it took.


If you would like an independent view of whether your firm would keep working through a day like Thursday, I offer a complimentary health check with no obligation. The contact page is an enquiry form, so please put "free health check" in the message subject and I will pick it up straight away.


Ashley Barwick, Fraxinus Business Associates. This post is general information and does not constitute legal advice. Fraxinus Business Associates is not authorised to provide legal advice.


Sources



 
 
 

Comments

Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page