Air traffic controllers monitor radar screens overlooking an airport runway.

AI in Air Traffic Control: What It Actually Does, and What the FAA Ground Stop Really Was

Key Takeaways

  • The nationwide ground stop on Friday 13 November 2026 was not caused by AI. The Federal Aviation Administration, the FAA, ordered it after a communications failure affecting all 22 air route traffic control centres. United States officials said there was no indication of a cyberattack or a threat to the country, and the company behind the FAA’s AI system said its technology was not responsible.
  • The ground stop lasted more than four hours. It was ordered at 3.54pm Eastern Time and lifted at about 8.15pm Eastern Time. More than 10,000 flights were delayed and more than 4,000 were cancelled, according to flight-tracking figures reported on the day, with some counts putting cancellations higher as the evening went on.
  • It was only the second nationwide ground stop of all departures since the attacks of 11 September 2001. The reason, a communications equipment failure, has become increasingly common, according to reporting on the incident.
  • The FAA’s AI system, called SMART, does not control aircraft. It predicts congestion, suggests routes and timings, and passes its recommendations to human specialists for review. Human controllers retain responsibility for keeping aircraft separated, and they are free to accept or reject SMART’s suggestions.
  • The timing is what caused the confusion. SMART was expanded to all 22 centres on 10 November 2026, three days before the ground stop. That proximity, plus the fact that SMART is an AI system, led some people to ask whether it had caused the failure. The evidence available so far points the other way: the failure was in communications, not in traffic-flow advice.
  • AI in air traffic control today is mostly advisory, experimental or in trials. Real, live uses include traffic-flow prediction, arrival sequencing, runway monitoring and demand forecasting. No AI system in live operation autonomously separates aircraft, and no AI system replaces a human controller.

What happened on Friday 13 November 2026

On Friday afternoon, 13 November 2026, the FAA ordered a nationwide ground stop for departing flights across the United States. The order was issued at 3.54pm Eastern Time, or 8.54pm GMT, and it applied to all airports in the country. Aircraft already in the air were allowed to land. Aircraft on the ground could not take off.

The FAA said the stop was ordered following a communication issue impacting all Air Route Traffic Control Centers, the 22 regional centres that guide aircraft en route. In a post on social media, the FAA described the problem as a temporary telecommunications outage. The full advisory read, in part: “Due to extraordinary circumstances and for reasons of safety. Attention all aircraft operators, by order of the Federal Aviation Command Center all airports are ground stopped.”

The stop was lifted at about 8.15pm Eastern Time, more than four hours after it was ordered, after the FAA began gradually lifting the order. Some flights, including Pacific-bound and transatlantic departures, were cleared earlier. By the time the stop was lifted, the disruption was already severe, and it continued into the evening as airlines tried to get aircraft and crews back into position. Several airlines cancelled their remaining flights for the night, and Air Canada cancelled all flights to and from the United States for the remainder of the evening.

The scale of the disruption is best described as a range, because the counts changed as the evening went on and different tracking services reported different figures. Al Jazeera, citing FlightAware, reported more than 10,000 flights delayed and more than 4,000 cancelled. Flightradar24 reported more than 4,700 flights cancelled and 10,000 delayed, with another 700 already cancelled for Saturday. A later report put the figures higher still, at more than 9,000 delayed and more than 6,000 cancelled. About 4,042 flights were in the air, heading to United States airports, when the ground stop was ordered, according to Flightradar24. This article does not present a single definitive total, because no single total was stable at the time of writing, and the different figures reflect different counting times and methods rather than a disagreement about the scale. The scale was very large, whichever count is used.

The disruption spread beyond the United States, forcing the cancellation of flights from Canada and Europe. Passengers waited at airports across the country, and the effects were felt for hours after the stop was lifted, because a ground stop does not just delay the flights that were due to depart during it. It displaces aircraft and crews, and the system takes time to recover.

It was only the second nationwide ground stop of all departures since 11 September 2001. That fact, reported by Reuters and others, is what makes the event historic. The first, on 11 September 2001, was a response to a terrorist attack. This one was a response to a communications failure. The contrast is important, and it is worth holding onto it throughout this article.

What caused the ground stop

The cause, as described by officials on the day, was a communications failure, not an AI failure, not a cyberattack, and not a failure of air traffic control itself in the sense of controllers making a mistake.

Transportation Secretary Sean Duffy said the outage was caused by overloaded switches and routers. In a press briefing after the stop was lifted, he said that at about 3.36pm Eastern Time, officials encountered inconsistency in data and communications, and he blamed the problem on the country’s old air traffic system. He said the switches and routers shut down when they began providing inconsistent data.

CBS News, citing multiple sources, reported that a software update to a communications network operated by L3Harris Technologies caused the system to reboot and crash, triggering the outage. The network connects air traffic control facilities across the United States. That report attributes the cause to a software update, but it is a report based on unnamed sources, not a confirmed finding by the FAA, and it should be read as such. The FAA said its technicians were investigating the issue, and United States authorities said they had yet to determine the cause of the telecommunications outage at the time the stop was lifted.

Homeland Security Secretary Markwayne Mullin told reporters there was no indication that the ground stop was triggered by a threat to the United States. The Department of Homeland Security said there were no signs the outage was caused by a cyberattack. The FBI did not believe a cyberattack or criminal activity caused the failure, according to Bloomberg, citing an official familiar with the matter. Those statements do not prove what did cause the failure, but they do rule out, on the evidence available so far, the two explanations that would have been most alarming: a deliberate attack or a hostile act.

The National Weather Service also said on Friday that it was experiencing nationwide internet and communications issues, which suggests the problem was wider than air traffic control alone, although the connection between the two has not been confirmed.

What is clear is that the failure affected communications, and that communications are the foundation of air traffic control. Controllers cannot guide aircraft if they cannot communicate with each other, with pilots, or with the systems that share flight data. When communications fail across all 22 centres at once, the safe response is to stop departures, which is what the FAA did. That is a precaution, not a punishment, and it is the correct response to a loss of the systems that controllers rely on.

For a wider discussion of how AI safety is assessed, and what it means for a system to be safe to use, see our guide to AI safety, what it means, the key risks and how to use AI.

Was AI involved? The short answer is no, on the evidence so far

Search interest in “ai faa” rose sharply in the week before the ground stop, reaching Breakout status in Google Trends, and searches for “air traffic control” rose by about 3,700 percent. That rise reflects a real question people were asking: was AI involved in the ground stop? The answer, on the evidence available so far, is no.

The reason people asked is understandable. Three days before the ground stop, on 10 November 2026, the FAA expanded its AI air traffic system, SMART, to all 22 air route traffic control centres. SMART had been piloted in Washington, D.C. airspace since 21 September 2026, and the expansion was announced as a major step. Three days later, the entire system stopped. The timing alone was enough to invite the question, and it was asked widely on social media and in search queries.

But timing is not causation, and the evidence points away from SMART as the cause.

Air Space Intelligence, the company behind SMART, said its technology was not responsible for the nationwide ground stop. That is a company statement, and it should be read as such, but it is consistent with what is known about what SMART does and what failed. SMART is a traffic-flow management system. It predicts congestion and suggests reroutes. It is not a radio system, a radar system or a telecommunications system. The failure on 13 November was described by the FAA as a telecommunications outage, and by the Transportation Secretary as overloaded switches and routers. Those are communications systems, not traffic-flow systems.

Homeland Security Secretary Mullin said officials were looking at SMART as part of their investigation, according to one account, which is the correct response to a question of this kind: look at everything, including the new system, before ruling anything out. But looking at a system is not the same as finding it at fault, and no official has said that SMART caused the failure. The FAA has not named a root cause, and until it does, the responsible statement is that there is no evidence that AI caused the ground stop, not that AI has been definitively cleared.

It is also worth noting that SMART was not new on 13 November. It had been operating in Washington, D.C. airspace for almost two months, and it had already been deployed at some other centres before the nationwide expansion was announced. If SMART had a fault that could cause a nationwide communications failure, it would be surprising for that fault to appear only on the day of the expansion, and to manifest as a loss of radio frequencies and radar at en-route centres, which are not systems SMART controls.

What SMART actually is

To understand why SMART is unlikely to have caused the ground stop, it helps to understand what SMART is and what it does.

SMART stands for Strategic Management of Airspace, Routes and Trajectories. It is a cloud-based AI tool built by Air Space Intelligence, a Boston startup, under a 12-year contract reported at about 875 million dollars. The contract covers two systems: Flow Management Data and Services, known as FMDS, which is the planned technology backbone of the FAA’s Air Traffic Control System Command Center, and SMART, which the Transportation Department described as an enhancement within FMDS.

SMART centralises more than 200 data streams, covering weather, flight paths, traffic flow, controller staffing and other factors, in one platform. Reporting on the nationwide SMART rollout confirms the 10 November expansion to all 22 centres and the FAA’s claim of 180 rerouted flights since 21 September. It uses that data to predict traffic congestion, suggest alternative routes, establish the best arrival order for aircraft, and recommend scheduling changes. The aim, according to the FAA, is to reduce delays, reduce fuel consumption and improve the travel experience for passengers.

Since its launch in Washington airspace on 21 September 2026, SMART has helped reroute 180 flights, according to the FAA, saving an average of 10 minutes and almost 80 gallons of fuel per flight. Those figures come from the FAA itself and have not been independently verified, and this article does not present them as independently confirmed. They are included here because they are the figures the FAA has published, and because they give a sense of the scale at which SMART has operated so far: 180 flights, not thousands, and minutes and gallons, not hours and tonnes.

The most important fact about SMART is what it does not do. It does not separate aircraft. It does not issue instructions to pilots. It does not control an aircraft. It does not replace a controller. At the September launch, FAA Administrator Bryan Bedford said SMART will not be used by controllers whose job it is to prevent collisions. Transportation Secretary Duffy said, in the same period, “We are not outsourcing aviation to AI.”

Instead, SMART’s recommendations are reviewed by a team of FAA specialists at a SMART centre in Washington, who send scheduling recommendations to the Command Center in Warrenton, Virginia, which passes them to air traffic facilities for local leadership to consider. Controllers retain the final decision on whether to accept SMART’s suggestions. In other words, SMART is an advisory system, with humans in the loop at every stage, and with local controllers free to accept or reject its advice.

That design is deliberate, and it is the reason the claim that SMART caused the ground stop is hard to sustain. A system that only suggests routes, and whose suggestions are reviewed by humans before they are passed on, does not have the ability to shut down communications across 22 centres. It does not control the switches and routers that failed. It does not control the radio frequencies or the radar. It is, in the FAA’s own description, a planning tool, not a control system.

SMART has also attracted criticism, and that criticism should be reported fairly. The National Air Traffic Controllers Association, the controllers’ union, said it was not involved in the design, testing or implementation of SMART and therefore could not assess the system’s effect on the air traffic control workforce. Democrats in the House of Representatives expressed concerns about the tool and sent a letter to the Government Accountability Office asking it to review the FAA’s SMART implementation strategy, performance metrics and safety risk management. Those concerns are about the process by which SMART was introduced, and about whether its benefits have been properly measured, not about whether it caused the ground stop. They are legitimate questions, and they deserve answers, but they are different questions from the one this article is answering.

How air traffic control actually works

To understand what AI can and cannot do in air traffic control, it helps to understand how air traffic control works without AI.

Air traffic control in the United States is organised in layers. At the top, in terms of geography, are the 22 Air Route Traffic Control Centers, known as ARTCCs, which guide aircraft en route, between airports. Below them are Terminal Radar Approach Control facilities, known as TRACONs, which handle aircraft arriving at and departing from busy airports. At the bottom are the control towers at individual airports, which handle aircraft on the ground and in the immediate vicinity of the airport.

All of these facilities are coordinated by the Air Traffic Control System Command Center in Warrenton, Virginia, which manages the flow of traffic across the whole system. That is the centre that issued the ground stop on 13 November, and it is the centre that SMART’s recommendations are passed to.

Controllers in each facility are responsible for keeping aircraft separated, which means keeping them a safe distance apart, both horizontally and vertically. That is the core safety function of air traffic control, and it is performed by humans, using radar, radio and flight data, with the support of automated systems that alert them to potential conflicts. No AI system in live operation in the United States, or anywhere else that this article could verify, has taken over that separation function.

The reason is not that AI could not, in principle, help with separation. It is that separation is a safety-critical function, and safety-critical functions are subject to far stricter requirements than advisory functions. A system that suggests a route can be wrong, and the error can be caught by a human. A system that separates aircraft cannot be wrong, or the consequences are immediate and severe. That difference explains why AI has been introduced first in advisory and planning roles, and why it has not been introduced in live separation.

Communications are the thread that holds the whole system together. Controllers need to talk to pilots, to each other, and to the systems that share flight data. When communications fail, as they did on 13 November, controllers lose the ability to perform their function safely, even if the aircraft themselves are perfectly capable of flying. That is why a communications failure leads to a ground stop, and why the ground stop was the correct response.

What AI actually does in air traffic control today

AI in air traffic control is real, but it is narrower and more modest than the headlines suggest. The live and tested uses fall into a few categories, and each one is worth describing separately.

Traffic-flow management: the FAA’s SMART

SMART, described above, is the most prominent live example in the United States. It predicts congestion, suggests routes and timings, and helps controllers and flow managers plan ahead. It is a traffic-flow management tool, not an air traffic control tool in the strict sense, because it does not separate aircraft. Its value, if the FAA’s figures are accurate, is in saving minutes and fuel, not in preventing collisions.

Arrival management: A(I)MAN in Germany

In Germany, the air navigation service provider DFS has developed an AI arrival manager, known as A(I)MAN, which uses machine learning to plan the sequence in which aircraft arrive at an airport, and to estimate their arrival times. The system uses a neural network for sequence calculation and a gradient boosting method for estimate calculation, according to a presentation by DFS at a EUROCONTROL FlyAI Forum in 2023. The system was tested and verified within a funding project on AI certification, and it is described as a controller assistance tool, not a replacement for a controller. Simulation runs with air traffic controllers showed good acceptance, and the arrival planning quality was equivalent to or better than a conventional rule-based system. That is a real, tested use of AI in arrival management, but it is an assistance tool, and it is subject to certification requirements that are still being developed.

Runway monitoring: AIMEE at Heathrow

At London Heathrow Airport, the air traffic management operator NATS has been trialling an AI system called AIMEE, developed by the Canadian company Searidge Technologies, to help reduce delays in low visibility. The system uses ultra-high-definition cameras and AI to track aircraft on the runway and to tell the controller when an aircraft has successfully cleared the runway. The controller then makes the decision to clear the next arrival. The system was tested on tens of thousands of arriving aircraft, with reports putting the figure at 40,000 flights in one account and more than 50,000 in another, and the findings were to be presented to the United Kingdom Civil Aviation Authority. AIMEE is an assistance tool. It does not clear aircraft to land or take off. It tells the controller what it has seen, and the controller decides.

Digital twins and autonomous control: Project Bluebird

In the United Kingdom, NATS, the Alan Turing Institute and the University of Exeter are working on Project Bluebird, a multi-year effort to create AI agents that could, in theory, autonomously control air traffic. At Heathrow, NATS has also been trialling an AI runway-monitoring system called AIMEE to tell controllers when an aircraft has cleared the runway. The project is building a digital twin of United Kingdom airspace, a real-time, physics-based computer model that predicts future flight trajectories, and training AI agents, sometimes called digital controllers, to manage that airspace within the twin. The plan is to run shadow-mode live trials, in which an AI agent is fed real-time information and sends instructions to digitally represented aircraft within the twin, while human controllers manage the real airspace, and the two are compared.

That is research, not live operation. The researchers themselves say that an autonomously functioning digital controller would take a very long time to regulate and get into operations, even if it were possible to automate the entire task, which they say is unlikely. Project Bluebird is therefore best understood as an experiment to find out how far AI can go, not as a system that is about to replace controllers.

Demand, capacity and sector design: research in Europe

A European research project called SMARTS, funded by the SESAR Joint Undertaking and involving Lancaster University, NATS, the German Aerospace Center, EUROCONTROL and others, is using machine learning, mathematical modelling and optimisation to redesign flexible airspace sectors, with the aim of reducing delays, unlocking shorter routes, lowering emissions and alleviating controller workloads. The project began at the end of 2023 and was due to run until 2026. That is research into how airspace could be organised more efficiently, not a live control system.

Taken together, these examples show a pattern. AI in air traffic control is being used to predict, to plan, to advise and to assist. It is not being used to separate aircraft, to issue instructions to pilots, or to replace human controllers. That pattern is likely to continue for some time, because the safety and certification requirements for a system that separates aircraft are far higher than those for a system that suggests a route.

What AI does not do in air traffic control

It is worth stating the limits plainly, because they are often lost in the excitement about AI.

AI does not fly the aircraft. Pilots fly aircraft, with the support of autopilot systems that have existed for decades, and those autopilot systems are not AI in the sense that SMART or AIMEE are AI. They are automated control systems, designed to follow a set course, not to learn or to make judgements.

AI does not separate aircraft in live operation. No system in live use has taken over the separation function from human controllers, and no regulator has certified such a system. The research projects that are exploring autonomous separation, such as Project Bluebird, are doing so in digital twins and simulations, not in live airspace.

AI does not control the radio, radar or telecommunications systems that air traffic control relies on. Those are infrastructure systems, and their failure, as on 13 November, is an infrastructure failure, not an AI failure. SMART does not run those systems, and it cannot shut them down.

AI does not make the decision to order a ground stop. That decision is made by humans, at the Command Center, on the basis of the information available to them. On 13 November, the decision was made because communications had failed, and it was the correct decision.

And AI does not, on its own, make air travel safer or less safe. It is a tool, and its effect depends on how it is designed, tested, certified and used. A well-designed advisory tool, with humans in the loop, can make the system more efficient. A poorly designed tool, or a tool introduced without proper consultation or testing, can create new risks, even if it does not cause a specific incident. The debate about SMART is a debate about that second point: whether it was introduced with enough consultation, testing and measurement. That debate is legitimate, and it should continue. It is not the same as saying SMART caused the ground stop.

Why the confusion happened

The confusion about AI and the ground stop is understandable, and it is worth explaining why it happened, rather than simply dismissing it.

First, the timing. SMART was expanded to all 22 centres on 10 November. The ground stop happened on 13 November. Three days is a short gap, and the human mind is quick to connect events that happen close together in time. That is a natural tendency, but it is not evidence, and it is the reason investigators look for a mechanism, not just a coincidence, before attributing a failure to a cause.

Second, the name. SMART is described as an AI tool, and it is an AI tool. But “AI” is a broad term, and it covers everything from a chatbot to a traffic-flow predictor to a hypothetical autonomous controller. When people hear that an AI system was rolled out three days before a nationwide ground stop, they may imagine a system that controls aircraft, rather than a system that suggests routes. The distinction between traffic-flow management and air traffic control is technical, and it is not widely understood outside aviation. That gap in understanding is what allowed the rumour to spread.

Third, the search data. Searches for “ai faa” reached Breakout status, and searches for “air traffic control” rose by about 3,700 percent. Those figures show that many people were asking the same question at the same time. That is not evidence that AI was involved. It is evidence that many people were worried that it might have been, and that they turned to search to find out. This article is an attempt to answer that question with the evidence available.

Fourth, the wider context. The ground stop happened at a time when AI is being introduced into many safety-critical systems, and when there is legitimate public debate about whether that introduction is happening too fast, with too little oversight. The SMART rollout itself was criticised by the controllers’ union and by lawmakers, for reasons that have nothing to do with the ground stop. That criticism created a background in which it was plausible, to many people, that an AI system had caused a failure. Plausibility is not proof, but it explains why the question was asked so widely.

The responsible response to that confusion is not to dismiss the question, but to answer it carefully, with the evidence, and to be clear about what is known and what is not. That is what this article has tried to do.

The old system, and why it keeps failing

The ground stop also raises a question that has nothing to do with AI, but everything to do with the state of American air traffic control: why does the old system keep failing?

Reuters, in its reporting on the ground stop, noted that the grounding of all departing flights on Friday was only the second such shutdown since 11 September 2001, but that the reason, communications equipment problems, has become increasingly common. That is a significant observation. It suggests that the ground stop was not a freak event, but the most severe manifestation of a pattern.

The pattern is visible in the reporting of the past three years. There have been a series of disruptions caused by ageing equipment, software failures and communications problems. The Transportation Secretary blamed the problem on the country’s old air traffic system, and he was referring to a system that, in some parts, still relies on technology that is decades old. The FAA has said it will finally rid itself of dependency on the 55-year-old floppy disk format by 2028, which gives a sense of how old some of the technology is.

That is the context in which SMART was introduced. The FAA’s argument is that AI can help manage a system that is under strain, by predicting congestion, suggesting routes and reducing delays. The critics’ argument is that AI cannot fix an ageing infrastructure, and that introducing a new AI system without fixing the underlying equipment is a distraction. Both arguments have merit, and the ground stop does not settle the debate. It does show, however, that the most urgent problem in American air traffic control is not AI. It is the age and reliability of the communications and equipment that controllers rely on every day.

What to watch next

Several things are worth watching as the investigation into the ground stop continues.

The first is the FAA’s final determination of the cause. The FAA said its technicians were investigating the issue, and the Transportation Secretary attributed it to overloaded switches and routers. CBS News reported, citing multiple sources, that a software update to an L3Harris-operated network caused the system to reboot and crash. Whether that report is confirmed, and whether the FAA publishes a full account of what happened, will determine how much we ultimately know about the cause. Until then, the cause should be described as a communications failure, with the specific trigger still under investigation.

The second is whether SMART is mentioned in that final account. If the FAA finds that SMART played no role, that will be an important finding, and it should be reported as such. If it finds that SMART played a role, however indirect, that too should be reported, and the implications should be examined. At the time of writing, there is no evidence that SMART played a role, and the company behind it has denied that it did.

The third is the response of the controllers’ union and of lawmakers. The union said it was not involved in the design, testing or implementation of SMART. Lawmakers asked the Government Accountability Office to review the SMART implementation. Whether those concerns lead to a pause, a review or a change in how SMART is used will be a significant development, and it will affect how AI is introduced into air traffic control in the future.

The fourth is the wider question of infrastructure. Will the ground stop lead to faster investment in the communications and equipment that failed, or will it be treated as a one-off event? The answer will tell us whether the system has learned from the failure, or whether it is waiting for the next one.

Finally, it is worth watching the language. If “AI caused the ground stop” continues to be repeated without evidence, the public will be misled about both AI and air traffic control. If the distinction between traffic-flow management and air traffic control is explained clearly, the public will be better equipped to judge future claims about AI in aviation. Accuracy about language is not a minor matter here. It is the difference between a public that understands a risk and a public that is frightened by a rumour.

For a related warning about AI and infrastructure, and the difference between a judgement and a forecast, see our report on Geoffrey Hinton’s AI blackout warning, and for the wider debate about AI safety, see our report on Kanishka Narayan’s AI safety warning.

What We Could Not Verify

  • The final, definitive cause of the communications failure. The Transportation Secretary attributed it to overloaded switches and routers, and CBS News reported, citing multiple sources, that a software update to an L3Harris-operated network caused the system to reboot and crash. The FAA said its technicians were investigating, and no final root cause had been confirmed at the time of writing. The software-update explanation is therefore reported, not confirmed.
  • A single definitive total for flights cancelled and delayed. Different tracking services reported different figures at different times: more than 4,000 cancelled and more than 10,000 delayed, more than 4,700 cancelled and 10,000 delayed, and more than 6,000 cancelled and more than 9,000 delayed. This article reports the range rather than a single total, because the counts changed as the evening went on.
  • The FAA’s figures for SMART’s performance. The FAA says SMART has helped reroute 180 flights since 21 September 2026, saving an average of 10 minutes and almost 80 gallons of fuel per flight. Those figures come from the FAA and have not been independently verified, and this article does not present them as independently confirmed.
  • Whether SMART was definitively cleared of any role in the ground stop. Air Space Intelligence said its technology was not responsible, and no official has said it was, but the FAA has not published a final determination, and Homeland Security officials said they were looking at SMART as part of their investigation. The correct statement is that there is no evidence SMART caused the failure, not that it has been definitively cleared.
  • The exact number of flights in the air when the ground stop was ordered. Flightradar24 reported 4,042 flights in the air, heading to United States airports. That figure comes from a single tracking service and has not been independently confirmed.
  • Whether the National Weather Service communications issues reported on the same day were connected to the FAA outage. The Weather Service said it was experiencing nationwide internet and communications issues, but the connection between the two has not been confirmed.
  • The precise scope of the L3Harris network involved, and the nature of the software update reported by CBS News. That report is based on unnamed sources, and L3Harris did not immediately respond to a request for comment, according to the reporting.

The bottom line

AI did not cause the FAA ground stop on 13 November 2026, on the evidence available so far. A communications failure did. The stop lasted more than four hours, delayed more than 10,000 flights and cancelled more than 4,000, and it was only the second nationwide ground stop since 11 September 2001.

AI in air traffic control is real, but it is advisory, not controlling. The FAA’s SMART system predicts congestion and suggests routes. Human specialists review its suggestions, and human controllers decide whether to accept them. In Germany, an AI arrival manager helps sequence aircraft. At Heathrow, an AI system helps controllers know when a runway is clear. In the United Kingdom, researchers are testing AI agents in a digital twin, not in live airspace.

None of those systems separates aircraft, issues instructions to pilots, or controls the communications network that failed on 13 November. That is the distinction to hold onto, and it is the reason the question “did AI cause the ground stop” can be answered, on the evidence so far, with a clear no.

Similar Posts