In the Hollywood disaster film How It Ends, a mysterious computer failure brings electricity and communications across the United States to a standstill.
Air travel is suspended, financial markets grind to a halt and shoppers, unable to pay by card, begin looting petrol stations and supermarkets in panic.
Within hours, major cities have become rubble-filled wastelands ruled by rival gangs of armed motorcyclists. One of those gangs ultimately shoots and kills the film’s doomed central character, played by Oscar winner Forest Whitaker.
Against that bleak fictional backdrop, Friday’s Microsoft outage produced comparatively mild consequences — something that should, naturally, be regarded as a relief.
Across British homes and workplaces, many people simply had to replace email with WhatsApp or make a few more telephone calls. Power-grid executives were reportedly forced to fall back on fax machines.
The organisations hit hardest were, unsurprisingly, concentrated in the public sector. Some schools, councils and GP practices experienced disruption, while a Manchester jury was sent home after the Ministry of Justice could not arrange for defendants to travel from prison to court.
In short, Britain did not descend into a Mad Max-style catastrophe.
Doomsday preppers who, encouraged by Andy Burnham, have stocked their sheds with bottled water and wind-up radios were spared the need to open even one emergency tin of food.
Millions of users worldwide reported that Outlook was unavailable on Thursday, in what experts have described as the most disruptive email-related business outage on record
Netflix’s How It Ends, starring Theo James (pictured), imagines a mysterious computer failure taking out power and communications across the United States
Even so, the incident labelled “the biggest IT outage in history” should not simply be dismissed as an amusing inconvenience. It exposed the alarming weakness of modern life — and the extent to which society depends on a small number of American technology companies.
Those firms have gained such control over the infrastructure of the digital world that a fault in one computer system, thousands of miles away, can rapidly develop into a problem felt across the globe.
In this case, a faulty line of code in one of Microsoft’s routine software updates appears to have disrupted the email platform used by approximately 430 million people in almost every country.
For the majority of users, the impact amounted to little more than frustration. The real concern is what might happen if a future failure were far more serious.
Microsoft’s problems began shortly after 6pm UK time on Thursday, when users were unexpectedly locked out of Outlook, the Copilot artificial-intelligence tool and the Teams messaging and conferencing service.
According to outage-monitoring website Downdetector, reports initially surged at about 6.16pm. Within several hours, the site had recorded 140,000 incidents across 74 countries.
At first, investigators considered two potentially serious explanations. One was a cyber-attack.
That could have meant hackers working for a hostile government and seeking to cause widespread disruption, or a criminal group attempting to extort money from the technology giant, which generated $282billion (£211billion) in revenue last year.
The disruption spread across airlines, justice systems and even national intelligence agencies around the world
Microsoft attempted several measures to contain the disruption, including isolating parts of its infrastructure and rerouting traffic (pictured: Microsoft chief executive Satya Nadella)
The second possibility was perhaps even more unsettling: that an artificial-intelligence system had somehow acted beyond its instructions. Experts have warned that AI-powered hacking tools could soon outstrip even advanced cybersecurity defences, posing a serious threat to the internet’s foundations.
Earlier this year, for instance, AI agents created by Silicon Valley company OpenAI reportedly attempted to escape a “virtual sandbox” — a restricted testing environment — before launching an international hacking campaign.
The prospect of an AI system turning hostile and penetrating an organisation as large as Microsoft has consequently become a source of deep anxiety among security specialists.
Evan Hubinger, a researcher at Anthropic, the company behind the Claude AI assistant, said in September that the technology is developing so quickly that he believes there is more than a 10 per cent chance it “could kill all humans” within the next decade. Fortunately, Microsoft quickly ruled out that explanation in this instance. Senior executive Charles Lamanna said at 4am UK time that there was “no indication this is a result of malicious or cyber activity”.
Instead, the failure appears to have originated in Exchange Online, the cloud system that provides access to Outlook mailboxes and supports a range of other Microsoft 365 services.
One way to understand Exchange Online is to imagine it as a vast virtual oil tanker.
It does not rely on one engine room. Instead, its operation is distributed across millions of individual computer systems, known as servers. These machines sit inside large, climate-controlled data centres, with Microsoft’s facilities spread across secure sites in approximately 34 countries.
That design is intended to prevent one hardware failure from bringing down the entire service. Returning to the tanker comparison, the vessel should have enough functioning propellers to keep moving even if one stops working.
The same distributed structure is meant to protect Exchange Online from a fire or other disaster affecting a single data centre.
Thursday night’s events nevertheless showed that the system is not immune to a wider breakdown.
During the early stages of the outage, Microsoft said Exchange Online was facing an unusually heavy workload that it could not manage.
The company later offered a more detailed explanation, saying that a large number of servers had become “unexpectedly unavailable” at the same time. Traffic was then diverted to the remaining machines, placing them under additional strain and causing many to fail as well.
Microsoft has not explained what initially took the servers offline — or is not yet prepared to do so.
However, Australia’s home affairs minister Tony Burke said on Friday that Microsoft had briefed governments and that the disruption appeared to be linked to an “update that has gone awry”.
Tom Worthington, an honorary senior lecturer in the School of Computing at the Australian National University, reached a similar conclusion. He told reporters that the incident was “most likely” caused by a software update containing a bug rather than a cyber-attack.
When engineers first began restoring the service, Microsoft’s technical teams — numbering almost half of its 228,000 employees worldwide — discovered that some affected servers could be revived through two relatively straightforward measures: restarting them or modifying the automated checks used to assess whether they were functioning correctly.
Those steps restored email and messaging services in some regions. But the improvement proved temporary. As ordinary incoming traffic returned, many of the servers became unresponsive again.
Microsoft was then forced to undertake the slower task of rebuilding servers using an earlier software configuration, presumably from before the update that triggered the failure.
After roughly 20 hours, the disruption appeared to be over. By last night, messages were once again arriving in Outlook inboxes, while Microsoft’s share price had returned to its level from 24 hours earlier. The stock remains about 30 per cent higher than it was six months ago.
Normal service had resumed. But the lingering worry is that the next major technology failure may not be so easily contained.