The FAA's AI went national. Three days later, flights stopped.
On Nov. 13 the FAA stopped flights nationwide for more than four hours, three days after taking its SMART AI routing tool to all 22 en-route centers. Officials blame overloaded routers, not the AI. What the timing teaches any company rolling out AI: how to prove what broke, and what to set up first.

In brief
On Friday, Nov. 13, the FAA halted flights across the U.S. for more than four hours. Three days earlier it had switched on its AI routing tool, SMART, at all 22 en-route air traffic centers. Officials say overloaded routers in old telecom equipment caused the stop, not the AI, and the investigation is still open. For a mid-size company the lesson isn't "AI breaks things." It's that when you roll out AI, it gets blamed for the next outage. Roll it out in stages, keep a person deciding, keep your own logs, and don't change two big things in the same week.
- The FAA stopped flights nationwide for over four hours on Nov. 13, three days after its SMART AI tool went live at all 22 en-route centers.
- Officials blame overloaded routers in aging telecom gear and say there's no evidence of AI interference; the investigation is ongoing.
- SMART only recommends; local managers decide. That design is why blaming it is hard.
- For any company: roll AI out in stages, keep a person deciding, keep your own logs, and don't stack big changes in one week.
What happened
Late Friday afternoon, Nov. 13, the FAA ordered a nationwide ground stop. According to AVweb, it started around 3:54 PM Eastern and was lifted at 8:15 PM. Transportation Secretary Sean Duffy said more than 3,500 aircraft were in the air when it was issued. At least 827 flights were diverted, a figure AVweb attributes to Cirium as reported by the AP. It was the first nationwide stop since January 2023.
The timing made headlines. On Nov. 10, three days earlier, the FAA said its new AI tool was active at all 22 air route traffic control centers, according to Flying Magazine. The tool is called SMART, short for Strategic Management of Airspace, Routes and Trajectories. It pulls in more than 200 data streams and recommends route changes.
So within hours, people were asking the obvious question: did the AI do this?
What officials say caused it
Officials say no. Per AVweb, they blamed overloaded switches and routers in aging telecom equipment, where older analog gear failed while newer digital systems kept running. Homeland Security Secretary Markwayne Mullin said investigators found no evidence of a cyberattack or AI interference. Duffy said, "There's no possible way SMART would have impacted what happened."
The company that built SMART, Air Space Intelligence, told Futurism the stop was not a result of its system. But the same report says Mullin, asked about SMART on live TV, said his agency was looking at it.
Independent experts didn't blame the AI either. In Scientific American, a former controller suggested a software update to the voice switching system may be to blame, and a George Mason University professor called the system brittle after years of cuts. Scientific American found no expert pointing at SMART. The only link is the calendar.
What's known, and from whom
| Claim | Who says it | Status |
|---|---|---|
| Overloaded routers in aging telecom gear caused the stop | Federal officials, per AVweb | Official explanation; investigation ongoing |
| No evidence of cyberattack or AI interference | DHS Secretary Mullin, per AVweb | Preliminary |
| The stop was not a result of SMART | Air Space Intelligence (the vendor), per Futurism | Vendor statement |
| A software update may have triggered it | Former controller Ruth Stilwell, per Scientific American | Expert opinion |
| SMART saved about 10 minutes and nearly 80 gallons per rerouted flight | FAA, per Flying Magazine | FAA's own figures, 180 flights |
Why the AI got blamed anyway
Because it was the newest thing in the room. That's not unfair; it's how people troubleshoot. When something breaks, you ask what changed. And SMART had arrived with critics. More than 20 members of Congress wrote to the Government Accountability Office on Oct. 13 asking for an independent review, as reported by KXAN. They said the frontline workforce had little or no say in how SMART was built.
Put those together and the story writes itself, whether or not it's true. A new AI system, people who weren't consulted, and then a bad day.
When you roll out AI, it becomes the first suspect for every problem that follows.
What this means for your office
You're not running air traffic control. But the same thing happens at a 150-person distributor. You turn on an AI tool that reads incoming POs. The next week, the ERP slows down, or an order ships wrong, or email goes quiet. The order desk will say it's the AI. Sometimes they'll be right.
The question is whether you can tell. Four things make that possible:
1. Roll it out in stages. The FAA did start small: according to AeroTime, SMART's pilot began Sept. 21 at the Washington-area airports before going national. Do the same. One customer, one desk or one document type first. Then widen it.
2. Keep a person deciding. SMART recommends; per AeroTime, local managers decide whether to act. That design is why "the AI grounded flights" is so hard to argue: it doesn't have the controls. Build your tools the same way. The AI drafts the order; a person approves it. We cover how in human-in-the-loop automation.
3. Keep your own logs. If your AI tool records what it read, what it did and when, you can answer "was it the AI?" in ten minutes instead of a week. If the vendor holds the only logs, you're taking their word for it, which is exactly where the FAA is with its contractor's statement.
4. Don't change two big things in the same week. If you go live with an AI tool and upgrade the ERP the same weekend, you'll never know which one caused Monday's problem. Space them out.
Where the other side has a point
The critics of SMART aren't wrong to ask questions. Lawmakers asked about testing, certification and training before the rollout. A national rollout in about seven weeks from the first pilot is fast for a safety system, and the people who use a tool every day should help shape it. That's true for an order desk too. If the people doing the work weren't asked, they'll trust the tool less and blame it more.
And the investigation isn't finished. If it turns out SMART played some part, the lessons above matter even more.
But the opposite mistake is real too. If every outage after a rollout gets pinned on the AI without evidence, companies will pull useful tools for the wrong reasons, and the real cause, here possibly old telecom gear, goes unfixed. As we saw with the Microsoft 365 outage the day before, the boring infrastructure under your AI tools is often what fails.
What to do this week
- List every AI tool your company turned on in the last six months, and who owns each one.
- For each, find out where its activity log lives and whether you can read it without calling the vendor.
- Check that each tool recommends or drafts, and a named person approves anything that reaches a customer or the ERP.
- Put your next go-live on a calendar that doesn't share a week with any other big change.
None of that slows AI down. It makes it something you can defend on the bad day.
Common questions
Did AI cause the FAA's Nov. 13 nationwide ground stop?
Officials say no. They blamed overloaded switches and routers in aging telecom equipment, and DHS Secretary Mullin said there was no evidence of AI interference. The investigation is still open.
What is the FAA's SMART system?
Strategic Management of Airspace, Routes and Trajectories, an AI tool that uses more than 200 data streams to recommend route changes. The FAA said on Nov. 10 that it was active at all 22 air route traffic control centers.
How long did the ground stop last?
From about 3:54 PM to 8:15 PM Eastern on Nov. 13, with at least 827 flights diverted, according to AVweb citing Cirium data reported by the AP.
Does SMART control flights by itself?
No. According to AeroTime, SMART staff send recommendations to the FAA Command Center and local managers decide whether to act on them.
How should a company roll out an AI tool safely?
In stages, with a person approving the AI's output, with logs the company can read itself, and without other big changes in the same week, so any problem can be traced.
Why do new AI tools get blamed for outages?
Because troubleshooting starts with what changed. A tool that was just rolled out, especially one the staff weren't consulted on, becomes the first suspect even without evidence.
Sources
- AVweb, FAA nationwide ground stop lifts, Nov. 13, 2026
- Flying Magazine, FAA SMART AI now active at 22 ARTCCs nationwide, Nov. 10, 2026
- AeroTime, FAA SMART air traffic tool expands to 22 en-route centers
- Futurism, U.S. airspace shuts down days after FAA rolled out AI, Nov. 13, 2026
- Scientific American, FAA grounds all U.S. flights days after expanding its AI air traffic system, Nov. 13-14, 2026
- KXAN via AOL, FAA groundings follow congressional warnings on AI, Nov. 14, 2026
Microsoft 365 was down for a day. Your AI went down with it.
Human-in-the-loop automation: why a person should approve every step
Why 95% of AI pilots fail, and why custom AI is the exception