Back

FAA Outage, November 13 2026: How One Router in Arkansas Halted Every US Departure

A contractor swapped a router at a small FAA radio site in Arkansas, and by afternoon no plane in the US could take off. Here’s what the FAA has confirmed, where the reports split, and the lesson for anyone whose vendors make changes for them.

Jagdish Patil

•

On Friday morning, November 13, a contractor installed a new router at a remote FAA radio site in El Dorado, Arkansas. By the afternoon, air traffic controllers were losing radio and radar feeds, and at 3:54 p.m. ET the FAA stopped every departure from every airport in the United States.

The explanation came the next day from Transportation Secretary Sean Duffy. A contractor was doing “routine maintenance to switch telecommunication lines from old copper wire,” he told reporters, according to Scripps News, and the new router started flooding the network with more than 1,000 connection attempts every four to five seconds.

That’s the headline. The details are messier. Sources don’t agree on how long the stop lasted, how many flights it cost, or whether anyone from the FAA was actually watching the work. This outage breakdown sticks to what’s confirmed and flags every place the reporting splits.

Incident Overview (TL;DR)

  • What failed: The private telecommunications network that links the FAA’s Air Route Traffic Control Centers. Controllers lost parts of their radio communications and radar feeds.

  • When it started: The FAA issued the nationwide ground stop at 3:54 p.m. ET (20:54 UTC) on Friday, November 13, 2026. Radio frequency problems were noticed hours earlier.

  • Where: Departures from every US airport. The router blamed for it sat at a remote radio site in Arkansas, and the FAA isolated a second one in Texas.

  • Scale: The FAA’s own questions and answers document, published Sunday, counted 12,462 cancellations across operators, 9,082 of them at major airlines, and said not all of them were caused directly by the ground stop.

  • Confirmed cause: A newly installed router, part of an L3Harris project moving FAA lines from copper to fiber, flooded the network with connection attempts. Officials say there’s no evidence of a cyberattack.

  • Status: Resolved. The stop lifted around 8:15 p.m. ET (01:15 UTC Saturday), and cancellations spilled into the weekend. The root cause investigation is still open, and no final incident report has been published.

What Happened

The FAA’s public wording was short. Its advisory cited “extraordinary circumstances and reasons for safety,” according to CNN. Domestic flights already in the air kept going to their destinations. Everything waiting at a gate stayed there. International arrivals were handled case by case.

What broke was the network underneath air traffic control, not the planes and not the airlines. CBS, as quoted on Wikipedia, described an update to “the private communications network that connects America’s air traffic control system” that caused the system to reboot. Duffy said network switches and routers were overloaded with data. The FAA later told Fox Business that it had isolated the Arkansas router and another in Texas, and switched off the broadcasting function on more than 60 similar routers.

The effects reached well past US borders. Eurocontrol held flights headed to the US from its member states at the FAA’s request. Mexico’s aviation authority kept aircraft bound for the US on the ground. Heathrow cancelled 29 flights to the US the next morning. Canadian systems weren’t affected, though flights into or over the US were.

Some early reports didn’t hold up. A Reuters item about a national weather radar outage was denied by the National Weather Service, which called it outdated content from isolated webpages.

Timeline

Times are approximate and in ET. Each line traces to the source named.

  • Monday, November 9: L3Harris notifies the FAA of a router replacement at a remote facility in El Dorado, Arkansas, scheduled for 9 a.m. Friday (CNN, citing sources).

  • Friday morning, November 13: The contractor carries out the installation (CNN).

  • Hours before the stop: Problems with radio frequencies are noticed (Wikipedia).

  • 3:54 p.m.: The FAA issues a nationwide ground stop for all departures (Wikipedia).

  • About 6 p.m.: Pacific flights out of San Francisco, including Hawaii and Asia, are cleared, along with some transatlantic departures from the East Coast (Wikipedia).

  • Friday evening: Eurocontrol suspends flights to the US from Europe at the FAA’s request (Wikipedia).

  • About 8:15 p.m.: The ground stop is lifted (Wikipedia).

  • Saturday, November 14: More than 2,500 additional flights are cancelled (AP, via Wikipedia). Duffy and FAA Administrator Bryan Bedford brief reporters on the router cause (Scripps News).

  • Sunday, November 15: The FAA publishes its questions and answers document with cancellation totals (Fox Business).

  • By Tuesday, November 17: The FAA pauses new router installations across its telecommunications hubs while it investigates (Fox Business, first reported by The Wall Street Journal). Bedford defends the oversight of the work at a press conference (CNN).

  • Wednesday, November 18: CNN reports sources saying “There was no monitoring” of the contractor’s work.

Where the Sources Disagree

How long it lasted. Wikipedia’s summary box says four hours and 21 minutes, from 3:54 to 8:15 p.m. ET. The same article’s body says three and a half hours, and gives a lift time of 01:19 UTC, which is 8:19 p.m. ET. CNN says departures were halted “for more than four hours.” We use roughly four hours twenty minutes.

How many flights. The FAA’s document counts 12,462 cancellations by operators, but says plainly that not all of them came from the ground stop. Wikipedia cites nearly 8,000 cancellations and more than 10,000 delays without naming the original source. AP counted more than 2,500 extra cancellations on Saturday alone. These cover different windows and definitions, so don’t quote any one as the total.

The 3,500 figure. Wikipedia says nearly 3,500 scheduled flights weren’t permitted to take off. CNN says about 3,500 planes were in the air when departures were halted. Same number, opposite meaning.

Who was watching. Bedford said the engineer “was monitored by an FAA technician” and that a risk assessment was done beforehand. CNN’s sources say otherwise: “There was no monitoring,” one said, and another said the technician “was only there to unlock the door.” PASS, the union for FAA technicians, said no member was asked to do a risk assessment. FAA spokeswoman Hannah Walden called the gap something that “could have been a misunderstanding or whatever.”

Which hardware. Wikipedia, citing The Wall Street Journal, says L3Harris was reportedly using Cisco routers. Other CNN coverage refers to Juniper router installations. Neither the FAA nor L3Harris has named a hardware vendor in anything we read.

Who Got Hit, and How

Travelers. Anyone booked on a departure Friday afternoon or evening sat at the gate or got cancelled, and the rebooking pushed the pain into Saturday. Flights already airborne landed normally.

Airlines and cargo. Wikipedia counts cargo flights in the disruption. The FAA’s document says operators cancelled many flights “out of an abundance of caution,” which is part of why its total runs so high.

International travel. Europe, Mexico and the UK all felt it. Flights that only crossed US airspace, including routes between Canada, Mexico, the Caribbean and Central America, were caught up too.

Businesses that run on travel. Wikipedia notes NBA, NHL, NFL and college football teams changed travel plans, and Amtrak saw a surge in phone and web traffic with some booking issues.

What wasn’t reported. The FAA expanded an AI traffic management tool called SMART to all its air centers on November 10, three days earlier, and plenty of people connected the two. The FAA and the tool’s developer, Air Space Intelligence, both denied it played a role, and Bedford said he saw no evidence of anything AI related either. The Microsoft 365 outage that began November 12 was also dismissed as unrelated. And if you saw airline apps spike on Downdetector that evening, remember it counts complaints, not affected customers. We found no report that airline systems failed on their own.

Root Cause: What the FAA Has Confirmed

Here’s the official record, with who said it:

  • Sean Duffy, Saturday press conference (Scripps News): a contractor was doing “routine maintenance to switch telecommunication lines from old copper wire” at “at least one of our remote FAA locations.” He said the new router sent more than 1,000 connection attempts every four to five seconds, and that there was no evidence of a cyberattack.

  • Duffy, to reporters (CNN): “We need L3Harris or whoever provided that router to us to do a better job.”

  • Bryan Bedford, FAA Administrator (Scripps News): “We see zero evidence of anything nefarious.” He also said he saw no evidence the AI tool was involved. Per Reuters, cited on Wikipedia, he also said the router explanation wasn’t a full explanation of the root cause.

  • The FAA (Fox Business): the router installed at a radio transmitter site in Arkansas and another in Texas were identified and isolated, and the broadcasting function was disabled on more than 60 similar routers.

  • L3Harris (Scripps News): said it has completed more than 70% of the planned move from copper lines to fiber, and is “supporting the FAA in its investigation to understand the conditions that led to the Nov. 13 ground stop.”

What nobody has said yet: why one router at a small site could swamp the whole national network, whether this was an installation mistake or faulty equipment, and why the work was scheduled on a busy Friday. Duffy framed the open question as “incompetence,” “user error or installer error from L3 Harris,” or a faulty router. Scripps reports investigators haven’t settled it.

Our Read of the Symptoms (Interpretation)

This section is our interpretation, not anything the FAA or L3Harris has confirmed.

When an entire national system stops at once, the usual suspects are an attack, a brand new system, or a shared dependency everything sits on. Here’s how each holds up.

A cyberattack. Duffy, Bedford and the White House all said there’s no evidence of one, and the physical trail leads to a specific router at a specific site. Weak fit.

The SMART AI tool. It went live everywhere three days earlier, which makes the timing tempting. But SMART suggests routes. It doesn’t carry radio and radar traffic, and the symptom described is a network drowning in connection attempts. Weak fit.

The shared network itself. This fits everything confirmed. Every Air Route Traffic Control Center depends on the same private telecom network, which CNN says L3Harris owns. One device at the edge started shouting, and the network apparently had nothing stopping that noise from reaching everyone else. The FAA’s own fix, switching off a broadcasting function on 60 more routers, points the same way: the problem was what the device was allowed to send, not where it sat. Strong fit.

There’s a process story under the technical one. CNN reports a March 2025 FAA memo that moved vendor telecom maintenance from requests to notifications, with the line “this work will proceed without FAA approval,” and only a handful of liaisons processing thousands of those notices. If that’s accurate, the router went in under a system designed to keep work moving, not to catch a bad change.

Context, not a link: this was one of several big failures that week, alongside Microsoft 365, AWS in Virginia, Chase and Apple. Nobody official has connected any of them to the FAA.

What This Means for Small Teams and Solo Founders

You don’t run air traffic control. But you almost certainly have the same failure in miniature: a vendor making a routine change to something you depend on, on a schedule you didn’t pick, announced by a notice nobody read. Your host’s maintenance email. Your DNS provider tweaking something. Your payment provider retiring an old API version.

The same week, our Chase outage breakdown traced a banking slowdown to a software release too. Changes cause most outages. A few things worth doing:

  • Treat vendor notices as change requests. When a host or DNS provider announces maintenance, put it on your calendar and decide who’s watching. A notice in an inbox is not oversight.

  • Don’t schedule risky changes for your busiest window. The FAA’s work ran on a Friday with heavy travel. For a store, that’s launch day. For an agency, it’s the day before a client demo.

  • Check right after every change, not hours later. The FAA had radio trouble well before the stop. Know within minutes whether your site still loads after you or a vendor touches it.

  • Limit the blast radius. If twenty client sites share one server, one bad plugin update can take all twenty down. Split the critical ones out.

  • Keep a way back. Know how to roll back a deploy, a DNS record or a config file before you change it.

How to Protect Your Site: Alarms, Not Notifications

That FAA memo summed up the problem in one word: notifications. A notification is something you can ignore. A notification that lands on a phone in Do Not Disturb is something you never see. If your site breaks after a vendor change while you’re asleep in another time zone, a silent push isn’t an alert. It’s a message you’ll read after your client already has.

That’s why we built SIOPS alarm first. When a monitor fails, SIOPS rings a Critical Alarm on your iPhone or Android that cuts through silent mode and Do Not Disturb, using iPhone Critical Alerts and the Android alarm channel. Every plan card lists call alerts, and alerts can also go out by email, push, Slack and webhook.

For a change like the one in Arkansas, set your checks up before the work starts: your homepage, your login, your checkout, and any API your app calls. Then you’ll know within one check interval of the change landing. The free Hobby plan covers 5 monitors on five minute checks, and the Serious plan checks every 60 seconds with unlimited alarms. Setup takes about two minutes.

One honest limit: SIOPS checks your public endpoints from the outside. It would never have seen the FAA’s private network filling up with connection attempts. What it sees is the moment that kind of failure reaches something your customers load.

FAA Outage FAQ

Was the FAA ground stop a cyberattack?

No, according to officials. Duffy said there was no evidence of a cyberattack, and Bedford said there was “zero evidence of anything nefarious.”

How long did the FAA ground stop last?

About four hours and twenty minutes, from 3:54 p.m. ET to roughly 8:15 p.m. ET on Friday, November 13. Some sources give slightly different end times, and cancellations continued into the weekend.

What caused the FAA outage?

A newly installed router at a remote radio site in Arkansas, put in by L3Harris during a copper to fiber upgrade, flooded the FAA’s communications network with connection attempts. Whether that came from an installation mistake or faulty equipment is still under investigation.

How many flights were cancelled?

The FAA counted 12,462 cancellations across operators, while noting not all were caused directly by the ground stop. Other tallies covering shorter windows are lower.

Did the FAA’s new AI system cause the ground stop?

The FAA and the developer of its SMART tool both denied it, and Bedford said he saw no evidence of AI involvement.

Are flights back to normal?

Yes. The ground stop lifted Friday evening and airlines worked through the backlog over the weekend. The FAA has paused new router installations while it investigates.

We’ll update this outage breakdown when the FAA publishes more about what went wrong.

Download SIOPS before your site is the next headline.

SIOPS rings a real alarm that cuts through Do Not Disturb, so you fix it before your client notices.

Similar Stories

Similar Stories