An incident response team reviews a network outage on a monitor

Six Outages in Six Days: Microsoft, the FAA, Chase, Apple and the NSA — Is AI to Blame?

Key Takeaways

  • Six major technology outages struck the United States between 12 and 17 November 2026, hitting Microsoft 365, the Federal Aviation Administration, JPMorgan Chase, United Airlines, Apple and the National Security Agency.
  • Officials have said there is zero evidence of cyberattacks and no evidence the incidents were related. Secretary of State Marco Rubio said on Tuesday that the government had no evidence the outages were connected.
  • Each organisation has named a separate technical cause: a latent software defect at Microsoft, aging telecom equipment at the FAA, a software bug at Chase, a technology issue at United, an internal update at Apple, and scheduled maintenance at the NSA.
  • The FAA's outage was caused by aging telecom gear overloaded with data — "1980s little copper lines", in Transportation Secretary Sean Duffy's words — and officials expressly ruled out any role for the FAA's new AI traffic-planning tool, SMART.
  • The debate is the story, not the answer. This article does not present the six outages as one connected event, and it does not present AI as their cause. It explains what happened, what officials say caused each failure, and why so many people are asking whether AI is to blame.

The week in order

It helps to see the six incidents as a timeline, because the timeline is what made the week feel like a single event — even though the evidence says it was not.

Thursday 12 November — Microsoft 365. Many business users lost access to Exchange Online email and other Microsoft 365 services. The disruption lasted 23 hours and 50 minutes, from 16:20 UTC on 12 November to 16:10 UTC on 13 November, according to Microsoft's service status. It was the first and longest of the six outages, and it set the tone for everything that followed.

Friday 13 November — the FAA. The Federal Aviation Administration ordered a nationwide ground stop, halting departures at every US airport. The stop lasted more than four hours — the longest since the 9/11 attacks — and disrupted air travel across the country. The FAA counted approximately 12,462 flights affected on Friday, including 9,082 cancellations by major airlines, and FlightAware recorded more than 9,900 cancellations on Friday alone.

Monday 16 November — Chase and United. Thousands of customers reported problems with online banking at JPMorgan Chase, Bank of America and US Bank, with Chase the most prominently affected. On the same day, United Airlines said a technology issue prevented it from sending pilots the final aircraft weight information needed before take-off.

Tuesday 17 November — Apple and the NSA. A large suite of Apple services went down, including the App Store, Apple Music, AppleCare, Apple Pay and iCloud sign-in, with Downdetector showing a spike in user complaints about Apple Pay. On the same day, the National Security Agency's internal network — NSANet, the top-secret system the agency uses for internal communications — went offline.

Six days, six incidents, six of the most important technology systems in the United States. It is not surprising that people connected them. It is also, on the evidence available, not correct to do so.

For the Microsoft incident in detail, see our earlier report on the Microsoft 365 outage of November 2026.

What caused each outage

Each organisation has given its own account, and the accounts are strikingly different from each other. That is itself evidence: a single cause would tend to produce a single story, and these are six different stories.

Microsoft: a latent software defect. Microsoft's preliminary post-incident report described three factors behind the 24-hour blackout: a latent software defect in a change whose deployment had begun in July; a shift in network traffic to a gateway that also supported Exchange Online; and roughly 14 percent of gateway capacity unavailable because of an unrelated issue. The picture is of a slow-moving failure — a bug introduced months earlier, activated by a traffic shift, and compounded by reduced capacity. Microsoft's final report was expected on 20 November, and the preliminary findings should be read as exactly that: preliminary.

The FAA: aging telecom equipment. The FAA's account is the most detailed of the six, and the most politically charged. Transportation Secretary Sean Duffy said the agency began seeing unstable radar and communications data at air traffic control facilities across the country at 12:36 p.m. Pacific on Friday. Aging switches and routers in the FAA's telecom network "got overloaded with data" and began producing inconsistent information, he said, adding that the agency still relies on "1980s little copper lines". His line at the White House briefing was blunt: "Our new IP switches did not go down. It's the old hardware that went down."

Homeland Security Secretary Markwayne Mullin, standing beside him, added the trigger: the overload appeared to have been caused by new equipment the FAA was installing, as the agency added patches and new technology that overwhelmed the older system. The FAA's initial investigation pointed to work at a radar site in Amarillo, Texas, and a radio site in El Dorado, Arkansas, where a contractor — L3Harris — was doing routine maintenance, replacing copper wire with fibre optics. The FAA restored service after deploying fixes to reduce the load on its switches, and airspace reopened at 5:15 p.m. Pacific, more than four hours after the problems began.

JPMorgan Chase: a software bug. The bank attributed its disruption to "a software bug in on-premise hardware systems". That is a notably old-fashioned explanation — on-premise hardware, not the cloud — and it is a reminder that the largest banks in the world still run critical systems on infrastructure they own and operate themselves.

United Airlines: a technology issue. United described its problem in the vaguest terms of any of the six, calling it simply "a technology issue" that prevented the delivery of final aircraft weight data to pilots. Weight and balance information is safety-critical — an aircraft cannot legally take off without it — so even a brief failure has immediate operational consequences.

Apple: an internal update. Apple said an internal update was behind the problems affecting iCloud, the App Store, Apple Music, AppleCare and Apple Pay. The company's system status page acknowledged full outages for Apple Music, the App Store and AppleCare, and issues with iCloud sign-in. Apple Pay, which is used by millions of people for everyday payments, showed a sharp spike in user complaints on Downdetector.

The NSA: scheduled maintenance. The NSA said its classified internal network went down during "scheduled routine maintenance". A spokesperson confirmed the agency had experienced "temporary degradation during scheduled routine maintenance", and Director of National Intelligence Jay Clayton linked the incident to security patching. That is the least alarming explanation of the six — maintenance, not failure — but it is also the one that attracted the most speculation, because the NSA's internal network is not a system the public expects to hear about at all.

What the officials said

The official response to the week has been consistent, emphatic and — for the moment — evidence-based. There is no evidence of cyberattacks, and no evidence the incidents were related.

Secretary of State Marco Rubio addressed the speculation directly on Tuesday in Manila, saying officials had no evidence that the incidents were related and that there was zero evidence of cyberattacks. Director of National Intelligence Jay Clayton said there was no indication of an attack. Homeland Security Secretary Markwayne Mullin said the FAA failure appeared mechanical, with no evidence of a cyberattack, and the FBI does not believe a cyberattack or criminal activity was behind the FAA failure, according to Bloomberg's reporting.

Those statements describe the evidence available at the time, not a final determination. Only Microsoft had published a technical report by 18 November, and that report was preliminary. The FAA's root cause analysis was still under way, with no final report published. The other organisations had given explanations but not detailed technical accounts. The officials' statements are therefore best read as a summary of what the evidence shows so far — which is no evidence of an attack — rather than as a guarantee that no attack will ever be found.

That distinction matters, because it is the gap in which speculation grows. "No evidence of an attack" is not the same as "proof there was no attack", and the public conversation about the week has been shaped by that ambiguity as much as by the facts.

The AI question

The question that has dominated the coverage is whether artificial intelligence caused the outages. It is a fair question to ask, and it has at least three different versions, which are often confused with each other.

Version one: did an AI system directly cause an outage? On the evidence available, the answer is no — and in the one case where an AI system was a plausible suspect, officials have expressly ruled it out. The FAA's new SMART tool — a cloud-based system that pulls in about 200 data streams for air traffic planning — was the obvious candidate for blame in the ground stop. Duffy ruled it out at the evening briefing, saying SMART is not connected to the core systems controllers use to track flights. Air Space Intelligence, the company that built SMART, also denied that its products or its FAA integration caused the ground stop. GeekWire, which covered the briefing in detail, headlined its report: "Officials blame aging FAA telecom gear as U.S. airspace reopens, see no link to Microsoft or AI."

Version two: did AI contribute indirectly, by adding load or complexity to the systems that failed? This is a more serious question, and it does not have a clean answer yet. AI workloads are consuming enormous amounts of computing power, network capacity and electricity, and they are being integrated into systems — including air traffic control, banking and cloud infrastructure — that were not designed for them. The FAA's failure was, on the official account, a story about old equipment being overwhelmed by new demands. That is not an AI story in the direct sense, but it is a story about the same underlying risk: systems being asked to do more than they were built for, with the new load arriving faster than the old infrastructure can be replaced.

Version three: is AI making infrastructure more fragile in general? This is the question behind headlines like "Microsoft, FAA outages spark warning of AI-fueled cascading failures" and "Mass outages… Is AI to blame?" It is a question about the direction of travel rather than about any single incident: as more systems become interconnected, automated and dependent on AI, does the risk of cascading failure increase? That is a legitimate debate, and it is one that infrastructure experts have been having for years. But it is a debate about risk, not a finding about cause. The six outages of this week do not prove that AI is making infrastructure more fragile, and they do not prove that it is not. They prove that six important systems failed in the same week for six different reasons, and that the public is — reasonably — asking whether that is a coincidence.

For a deeper look at how AI is being built into the systems that failed, see our coverage of Microsoft's AI models for coding and cybersecurity.

Why the week felt like one event

There are good reasons the six outages felt connected, and they are worth taking seriously rather than dismissing as paranoia.

First, the timing. Six major failures in six days is a striking cluster, and human beings are pattern-seeking animals. When the email goes down on Thursday, the planes stop on Friday, the banks falter on Monday and the NSA goes dark on Tuesday, the natural response is to look for a single explanation.

Second, the opacity. Most of the organisations involved gave brief, technical explanations that left large gaps. Microsoft's report was preliminary. The FAA's investigation was ongoing. United's explanation was a single phrase. When official accounts are thin, speculation fills the space — and the speculation that fills it fastest is usually the most dramatic one.

Third, the context. The week arrived in the middle of a genuine debate about AI and infrastructure. The FBI had just reissued a warning about AI-powered cyberattacks. California was suing OpenAI over allegedly rogue AI agents. Google's own Gemini agents had reportedly gone rogue in a safety test. In that environment, the idea that AI might be behind a wave of outages did not sound absurd; it sounded like the next chapter of a story people were already following.

And fourth, the stakes. The systems that failed are the systems modern life depends on: email, air travel, banking, payments, government communications. When they fail together, the question "is this an attack?" is not a conspiracy theory. It is a rational first question, and it deserves a better answer than "trust us".

The real lesson: maintenance, age and change

If there is a common thread in the six outages, it is not AI and it is not an attack. It is something more mundane and more worrying: the gap between the systems we depend on and the state those systems are actually in.

Look at the causes again. A software defect introduced in July and discovered in November. Telecom equipment from the 1980s, still carrying air traffic control data, being overwhelmed by new equipment installed alongside it. A software bug in on-premise bank hardware. An internal update that took down a suite of consumer services. Scheduled maintenance on a classified network. What these have in common is not a single villain but a single pattern: complex systems being changed while they are running, on infrastructure that is older than the people maintaining it, with the consequences of a mistake propagating faster than anyone can contain them.

The FAA's case is the clearest illustration. The agency is in the middle of replacing its legacy telecom network with a modern internet-protocol network. AT&T took over the lead role on that project in August with an initial $74.3 million award, and the FAA aims to finish by the end of 2028. Every month until then, old and new hardware run side by side — and the November failure happened precisely at the seam between them, when new equipment overwhelmed old switches. The Government Accountability Office had already warned in September that the FAA had never built an integrated master schedule for the first phase of its air traffic control rebuild. The audit did not predict a nationwide switch overload, and nobody should claim it did. But it did describe a programme being run without the planning that a programme of its size requires.

Microsoft's case is similar in structure. A change deployed in July contained a latent defect. It took months, and a shift in network traffic, for the defect to bite. That is a story about the difficulty of testing changes to systems at planetary scale — and about how the cost of a mistake grows with the size of the system.

The uncomfortable conclusion is that the six outages may not share a cause, but they may share a condition: a technology infrastructure that is being asked to do more, changed more often, and depended on more heavily than ever before, while the work of maintaining and replacing it lags behind. That is not as satisfying as a single explanation. It is, on the evidence, more likely to be true.

What happens next

Several things are still to come. Microsoft's final report was expected on 20 November, and it may revise or deepen the preliminary findings. The FAA's root cause analysis is still under way, and a final report will be the first authoritative account of what happened to the telecom network. The other organisations may publish more detail, or they may not — banks and government agencies are not known for volunteering technical post-mortems.

The debate about AI's role will continue regardless. The question is whether it continues as a debate about evidence — what AI systems actually did, in which incidents, with what consequences — or as a debate about vibes, in which every outage is attributed to AI because AI is the thing everyone is worried about. The first kind of debate is useful. The second is not, and it actively harms the first by making it harder to distinguish real risks from imagined ones.

For readers, the practical takeaway is narrower than the headlines suggest. The six outages were real, they were disruptive, and they exposed genuine fragility in systems that matter. But the evidence available today points to separate technical causes, not a coordinated attack and not an AI failure. That could change as investigations conclude, and if it does, this article's conclusion should change with it. For now, the honest summary is the one the officials gave: no evidence of an attack, no evidence of a connection, and six separate problems that happened to arrive in the same week.

What We Could Not Verify

  • Final technical reports. Only Microsoft had published a technical report by 18 November, and it was preliminary. The FAA's root cause analysis was still under way, and the other organisations had given explanations but not detailed technical accounts. The causes stated in this article are the official preliminary explanations, not final findings.
  • Bank of America and US Bank. Some reports mention problems at Bank of America and US Bank on 16 November alongside Chase. The scale and cause of any disruption at those banks could not be verified from the material reviewed.
  • The full FAA flight figures. The FAA counted approximately 12,462 flights affected and 9,082 cancellations by major airlines, while FlightAware recorded more than 9,900 cancellations on Friday alone. These figures come from different counting methods and should not be added together.
  • The NSA outage's scope. The NSA confirmed temporary degradation of its internal network during scheduled maintenance, but the scope, duration and operational impact of the outage were not publicly detailed.
  • Any AI role. No evidence links AI to any of the six outages. That is a statement about the evidence available, not a proof that no such link exists — but no credible evidence of one had been published at the time of writing.
  • Microsoft's final report. A final report was expected on 20 November, after this article was written. Its findings may revise the preliminary account.

Sources: Firstpost; GeekWire (Todd Bishop, 13 November 2026); Press Report; Strong Mocha; Good Sidekick; FAA statement on the nationwide ground stop; Microsoft preliminary post-incident report as reported 12–18 November 2026; Forbes; Cybernews; Bergen Record.

Similar Posts