When Machines Advance and Humans Fall Behind (1/2): Lessons from the Shutdown of American Airspace

Friday, November 13, 2026, shortly after 3:30 p.m. in Washington. In Rabat it is 8:30 p.m., the time when I usually watch the international news channels to keep up with what is happening in the world. Suddenly, a huge red “Breaking News” banner fills the CNN screen. The images stop me cold: dozens of aircraft stuck on the ground, and apparently at every airport in the United States. American airspace is closed. Not a storm, not an attack. An outage. And on a Friday the 13th, no less. The superstitious will no doubt see an omen in it. The engineer in me sees something else: luck has nothing to do with it, but the calendar will matter.

I was stunned. This is the country of Boeing, NASA, Silicon Valley and SpaceX! The largest airspace in the world, tens of thousands of flights a day. And that day, it took only a few minutes for everything to stop.

The cause? A router. A brand-new piece of equipment, installed in the middle of the afternoon at a remote site in Arkansas, which started flooding a network far too old to cope with the traffic.

I have spent a large part of my life in air transport: at Royal Air Maroc, at the head of two airlines I was fortunate enough to found, Air Sénégal International and Atlas Blue, and then at ONDA, Morocco’s airports authority. In this business, you learn one thing very early: disasters rarely come from a single big mistake. They come from small decisions nobody thought were important.

That is why this outage struck me so deeply. At heart, it is less the story of a failing technology than that of people and organizations who, by relying on machines and on their contractors, are gradually losing control of their own systems.

And the question I ask myself, and that I put to you, is simple: how could such a massive outage happen in such a technologically advanced country?

Five hours that paralyzed American airspace

Let us retrace that afternoon, as the FAA itself described it a few days later.

Around 3:30 p.m. Eastern time, the telecommunications network linking US air traffic control centers starts receiving an abnormal flood of data: nearly 1,000 connection attempts every four to five seconds. The oldest equipment, still built on technology inherited from copper lines, cannot withstand it. Communications with aircraft become unstable, and so do radar images.

Within minutes, 17 of the 22 major en route control centers are affected, followed by several approach facilities. The FAA then takes the only reasonable decision: it orders a nationwide ground stop. No more aircraft take off. Those already airborne, about 3,500 of them, are guided down using the frequencies and radars still working. I want to pay tribute here to the controllers and pilots, two professions for which I have always had the greatest respect: not a single incident, not a single injury, despite a true operational nightmare.

Where did this flood come from? The investigation points to two remote sites: a radar station in Amarillo, Texas, and a radio communications site in El Dorado, Arkansas. At the latter, the FAA’s contractor, L3Harris, had just installed a new router. Both sites are disconnected, the network settles down, and as a precaution the broadcast function is disabled on some sixty similar routers. To a network engineer, the scenario looks like a “broadcast storm”: a device flooding the network with messages addressed to all its neighbors until the network saturates. It could easily have been mistaken for a cyberattack, but US authorities say that hypothesis has been ruled out. Traffic resumes around 8:15 p.m.

Five hours of outage, but days of disruption: more than 12,000 flights cancelled on that Friday alone, and thousands of passengers stranded for the whole weekend.

As if that were not enough, the following Monday, United Airlines suffered its own IT outage, short but widespread, delaying departures at several airports. No link has been established with the FAA outage. But for travelers, the message is the same: the system is fragile.

In Washington, the debate quickly turned political. The Secretary of Transportation is asking for $50 billion to modernize air traffic control. But before talking about money, we need to ask how such an ordinary intervention could cause such chaos.

Three failures, none of them purely technical

When I read the FAA’s explanations, three things struck me. All of them come down to common sense, the kind of questions any airport or airline asks before touching a critical system.

The wrong day, at the wrong time. A Friday afternoon. For anyone who knows aviation, it is almost a reflex: you do not touch critical infrastructure on the eve of a weekend. Traffic is at its peak, teams are starting to leave, on-call staffing is reduced. If something goes wrong, you pay a heavy price. In the shoes of those in charge, I would have chosen midweek, in a low-traffic window, with every team on deck. The FAA has since acknowledged that it will review the timing of this kind of work. To me, that is an admission.

A routine nobody analyzes anymore. In our business, risk management is second nature. Every change, however modest, is assessed on two axes: the probability that it goes wrong, and the severity of what would happen if it did. So how could such an intervention slip through the net? My hypothesis comes down to one word: familiarity. Replacements like this have been carried out for decades, and they have always gone well. After so much success, you end up forgetting that the probability of a problem may be low, but its severity remains maximal. The contractor saw a router. Nobody, it seems, saw the airspace of an entire country behind it.

No safeguard between the new and the old. This is what surprises me most. When you connect modern equipment to an old network, you put protections at the junction: rate limiting, message filtering, network segmentation so that a local incident stays local. Here, a single router in Arkansas was able to destabilize 17 of 22 control centers. And the same function had to be disabled, as a precaution, on some sixty routers. Everything suggests that this equipment had been deployed without being properly tested under the conditions of the old network. The investigation will confirm it, or not.

None of these three failures is primarily about technology. They are failures of method, organization and vigilance. The machines did not fail on their own: it was people who stopped watching over them, and stopped anticipating the risks that could follow.

The deeper causes

Stopping at these three failures would be too easy. Behind them lie deeper causes, found far beyond the United States.

A system that ages while running. I have often used an image with my colleagues. Modernizing an airport, an airline or any infrastructure that must stay in operation is like operating on a patient to replace organs: the ultimate goal is that the patient stays alive, during and after the operation. You can transplant a heart, a lung or a liver in perfect condition, but if the patient dies on the table, you have failed. A system that stops dead, like American airspace that Friday, is exactly that failure. And everything you did, however brilliant and difficult, was for nothing. You cannot close American airspace to rebuild it. So you modernize piece by piece: a router here, a radar there, keeping the rest. The new piles up on the old, and it is precisely at the junction between the two that things break. Add budgets approved piecemeal, and you get modernization programs that drag on for decades.

The paradox of invisible maintenance. Aging systems have a twin, even less visible: upkeep that keeps getting postponed. In many countries, especially developing ones, people are happy to invest in the new: a terminal, a radar, a state-of-the-art system. These projects are visible; they get inaugurated. But when the time comes to sign a maintenance contract, expensive and renewed every year, financial oversight bodies hesitate. I know this from experience: I have had to defend, and get approved, very large budgets for maintenance contracts. I had to find allies and convince the members of the decision-making bodies one by one. It is never easy, because as long as everything works, these expenses look useless. It is on the day of the outage that you discover their real value.

Contractors who execute, an authority that no longer controls. It was a private contractor, L3Harris, that was working on the FAA’s network. I have nothing against contractors: aviation has always worked with them, and I have selected and managed many myself. But a contractor fulfils a contract, one job at a time. The public authority, on the other hand, carries the risk for the whole system. If it lets its in-house expertise go, it can no longer really oversee what it outsources, or say no when it must. The danger is not outsourcing. It is the loss of competence of the one who outsources.

Artificial intelligence and the illusion of verification. Let me say it up front: nothing published so far suggests that AI played any role in this outage. But it raises a question for tomorrow. More and more engineers ask general-purpose AI assistants to write a configuration or validate a procedure. These tools answer quickly, confidently, and often correctly. But they know nothing of the weaknesses of an old network in Arkansas. As Yann LeCun, Turing Award winner and one of the fathers of modern artificial intelligence, points out, these models have no real understanding of the physical world around them. A fluent answer gives the impression that you have checked, when you have only delegated. And the more you delegate, the less you know how to do yourself.

A series of outages, the same scenario

One might take comfort in thinking that November 13 was an isolated accident. Unfortunately, it was not. In recent years, I have seen the same scenario repeat itself, in different forms and in different sectors.

First, in the United States. In January 2023, the system that distributes notices to air missions, the NOTAMs, went down: departures had to be halted across the country. It was the first nationwide halt of departures since September 11, 2001. The cause? Contractor staff had deleted files by mistake while trying to synchronize the main database with its backup. A contractor again, and once more a job considered routine. In spring 2025, controllers at the Philadelphia facility handling the Newark approach lost radar and radio contact with aircraft for nearly 90 seconds, on April 28 and again on May 9. Seconds that, in a cockpit or in front of a radar screen, feel like an eternity.

Then in Europe. In March 2025, a fire at an electrical substation near Heathrow was enough to close Europe’s busiest airport for almost an entire day. A single power supply point, and some 200,000 passengers left stranded. A telling detail: according to the UK Energy Secretary, the fire also appears to have knocked out a backup generator. A month later, on April 28, 2025, the entire Iberian Peninsula was plunged into darkness: trains and metros stopped within seconds, and airports held up only thanks to their backup systems, at the cost of many delays and cancellations.

And at home in Morocco, on November 11, 2026, two days before the American outage, Casablanca’s power grid went dark in turn. I will come back to this.

The tech giants are not immune either. On November 12, 2026, the day before the American outage, Microsoft suffered a global outage: Outlook, Teams and even its AI assistant Copilot were down in more than 70 countries for over 24 hours, due to “high load” on its servers. The following Monday, even on Wall Street, JPMorgan Chase blamed a “software bug” after customers lost access to their accounts. Two days later, Apple’s App Store went down.

These outages lead me to a question that worries me more. Today, an Outlook outage means a day without email. But tomorrow? Our companies, our administrations, our hospitals, and perhaps one day our airports, rely more and more on a handful of artificial intelligence players: Nvidia for chips; OpenAI, Google with Gemini and Anthropic with Claude for models. What will happen the day one of them suffers an outage of this scale? If we have entrusted these tools with part of our thinking, our decisions and our operations, will we still be able to manage without them, even for a few hours? This is exactly the manual-mode question every airport asks about its critical systems, applied to our collective intelligence.

These outages do not share the same cause. A database, a radar, a transformer, a power grid. But they all look alike: a single point of failure, a redundancy that was not really one, and an organization that only discovers its fragility on the day of the outage. The technology changes from one case to the next. What was missing, every time, was control.

So, could this happen here? And above all, how do we make sure that, the day the system goes down, there are still men and women capable of taking back control? That is what I will address in the second part of this article: training those who will run our systems tomorrow, what Morocco has already been through, and the very concrete lessons risk management can teach us.


Zouhair Mohammed El Aoufir — Founder of Flying Tek, former senior executive in aviation and airport infrastructure.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *