The US aviation outage that halted takeoffs across the country this Friday (the 13th) turned into fuel for speculation within hours. The question going around was blunt: did the FAA AI play a part? Its national expansion had been announced just three days earlier. According to the Falls Church News-Press, there is no verified proof of any link. Even so, the episode brought back the warnings that Representative Don Beyer had been raising since September.
The News-Press cites the Associated Press, which reported that communication problems hit all 22 regional air traffic control centers. That led the FAA, the agency in charge of US civil aviation, to stop takeoffs nationwide. Transportation Secretary Sean Duffy said technicians were looking into it. As of the time the story was published, no official had named a cause.
In other words, what we know so far is that there was a communication failure. There is no official record of an artificial intelligence failure. Everything else, for now, is speculation.
What SMART is, and why it ended up under the spotlight
According to the outlet, which relies on information from the FAA itself, SMART runs in the cloud and pulls together about 200 data sources, such as weather, routes, traffic flow and controller schedules. The goal is to predict congestion and suggest changes to operations. It was designed to help with traffic management and flight scheduling. It does not fly planes, and it does not replace the conversation between pilots and controllers.
The FAA presents the tool as part of an effort to cut delays, reduce fuel use and ease the strain on American air infrastructure. That detail about what the system actually does matters a lot when sizing up the rumors. The timing alone proves nothing, and the News-Press makes a point of saying so.
What Don Beyer was asking the FAA for
Beyer co-chairs the bipartisan artificial intelligence caucus in the US Congress. According to the outlet, he called for SMART to be suspended on September 21. At the time, he said the people in his district should not be guinea pigs for an air traffic control system that, in his view, was still poorly tested. He also warned that rushing a technology tied to flight safety, without enough review, could put passengers at risk.
The congressman also said he had been told that controllers were not consulted during development and were not trained before the tool arrived at airports in the region. He criticized the lack of communication with the public too, especially after the deadly midair collision near Reagan National Airport in January 2025.
On October 9, Beyer and other lawmakers sent a letter to Duffy and FAA Administrator Bryan Bedford asking for a pause in the project. Their questions were very concrete. Was there independent safety testing? How does the system behave in severe storms or during system outages? What criteria would the FAA use to decide the technology was ready for the whole country? And who is accountable if an AI recommendation contributes to an incident?
This kind of pressure from lawmakers over technology is nothing new there, as we covered in how artificial intelligence is pressuring legislation in the US Congress.
The timeline question
The calendar raised eyebrows as well. Again according to the News-Press, on November 10 the FAA announced the expansion of SMART to all 22 regional centers, roughly 50 days after its debut in the Washington airspace. The problem is that the letters from Congress mentioned a 90-day review period. The outage came three days after the announcement.
That is why the outlet raises a question that is still unanswered: did the agency finish the planned reviews and hand lawmakers the safety studies they asked for? That question stands no matter what turns out to have caused the outage.
What this teaches about AI adoption in business
Not every company manages airplanes, but plenty of businesses in Brazil already use AI in processes that cannot fail: billing, inventory, customer service, logistics. Rushed rollouts, thin training and no plan for when everything goes down are risks in any project. They just get bigger when the system is one that cannot stop.
The simplest lesson is to test patiently before scaling up and to respect the review period you set for yourself. Picture a drugstore chain that uses AI to predict restocking at one store and, before measuring the results, rolls the tool out to every location. If something goes wrong, the problem shows up everywhere at once. And it becomes much harder to figure out where it started.
Another point is to listen to and train the people who will use the tool every day. They are the first to notice when a suggestion from the machine makes no sense. If they were left out of the process, they tend to either trust it too much or ignore it, and neither outcome is good.
Some of the questions the congressional letter put to the FAA work well for any team:
- What happens to operations if the system goes offline?
- Who is accountable for a decision suggested by AI?
- What criterion says the tool is ready to grow?
Answering these before launch is part of any serious approach to AI adoption in business.
Then there is communication. When clear information is missing, rumors fill the gap, and the FAA case shows that in real time. We dig into this effect in fake news, artificial intelligence and discernment in journalism.
The investigation is still open. The News-Press says it reached out to the office of Beyer to ask whether his concerns were addressed and notes that the story will be updated. Things may still change. In the meantime, it is worth checking out what military artificial intelligence teaches companies about using AI where mistakes are costly.
