- Ground stop
- Active
- Planned end
- 10 p.m. CT
- Root cause
- Unnamed; symptom: comms
The ground stop. How the story formed.
On a Friday afternoon the FAA stopped every departure in the United States and, for more than four hours, did not say why. We briefed clients through it in a run of seventeen timed updates, sixteen of them sent. This is the record of how the public story moved, how our brief moved beside it, and the seven errors we made along the way.
Every departure held. No cause given.
At 2:54 p.m. Central on 13 November 2026, the FAA’s Command Center halted departures at every U.S. airport, citing “extraordinary circumstances.” Flights already airborne continued to their destinations. Eurocontrol, at the FAA’s request, held U.S.-bound departures from its 42 member states overnight. The stop was lifted at about 7:16 p.m., roughly four hours and twenty minutes later. About 9,100 U.S. flights were canceled that day, nearly 6,700 on Saturday and more than 2,300 on Sunday, roughly 18,000 in all, as aircraft and crews were repositioned.
At a White House briefing after the lift, the Transportation Secretary called it “a legacy system problem.” Switches and routers became overloaded with data and, as he corrected himself, did not go down but “started to give us inconsistency in the data,” leaving communications and radar intermittent at the centers. Systems already upgraded from analog to digital, he said, did not have the problem, and new AI-powered systems had nothing to do with it. “What went down was old,” he said. “It’s going to continue to fail.” He asked Congress for $50 billion more for modernization.
Beside him, the Homeland Security Secretary framed it differently: “as Secretary Duffy and FAA was going in and putting patches into the system, they were putting new equipment and new technology and that new equipment and new technology … was overwhelming the older system.” He said there were no “footprints” of an intrusion, and added: “we’re still looking, so we’re not going to rule that out.” Neither named the change, the vendor, or confirmed the reported network update.
The two accounts can both be true: new technology can overwhelm old equipment that then fails. But one pointed at a change and the other at age, and that was the question the next stories would form around. Coverage leaned toward the age account.
Saturday, the trigger. The next morning the Transportation Secretary named the contractor and the work. L3Harris, a longtime FAA contractor, was doing what he called routine maintenance at a remote site, replacing copper with fiber, when a router overloaded a critical communication system. The FAA’s acting chief technology officer said two places were “pinging our network at a very high rate, a thousand connections over every four to five seconds”; the FAA disabled both and began a root cause analysis. The Secretary asked whether it was installer error or “a faulty router.” The work ran on a Friday afternoon; remote switchovers, he said, happen during the day, and there would be a “reevaluation of when we do this work.” The FAA Administrator called it “an equipment issue” and said there was “zero evidence of anything nefarious and zero evidence of anything AI-related.” He also narrowed the scope: as many as 17 of the 22 en route centers were affected, frequencies were “going in and out” and forcing switches between primary and backup, and none of the facilities “actually went dark.” L3Harris said it takes the stop and its impact on travelers seriously and is supporting the investigation.
Sunday, the sites. The FAA’s written account and the Transportation Secretary named the locations. The copper-to-fiber work, including a new router, was at a radio communications site in El Dorado, Arkansas. When that router was activated, equipment at a radar site in Amarillo, Texas, replaced about three weeks earlier, also began flooding the network. Blocking the two locations stabilized it, and broadcasting on the new routers was disabled as a precaution. Both routers went to the L3Harris lab in Florida for testing, with FAA, Juniper and Cisco engineers assisting. Hewlett Packard Enterprise, which owns router maker Juniper, said the device “did not fail or act improperly.” The FAA Administrator put the cost at 12,000 flights. The President blamed the previous administration while praising the FAA’s work and backing the funding request.
Monday and Tuesday, the equipment. The Secretary called the Arkansas router the “likely cause,” said FAA staff were examining it and the installation process at the L3Harris lab, and asked the FBI Director to send investigators. “We think it was the installation or the router that was the bad actor,” he said, while repeating that a cyberattack was not believed to be the cause. On Tuesday the FAA paused installation of new Juniper routers. Thirteen Democrats on the Senate Commerce Committee had asked for that pause, citing L3Harris figures: 3,648 routers planned, 400 installed, 69 operational. They asked why the backups failed to contain the disruption, noted that the early accounts from the Department and from L3Harris differed, and requested a joint briefing with the FAA, L3Harris and Juniper by 23 November.
Saturday refined the Friday threads rather than replacing them. CBS’s network update that crashed on reboot became a physical change to the same network; both secretaries’ framings turned out partly right, new work on old infrastructure; and the brief’s expectation, a change to one shared network reaching most of the centers at once, held. The event resolves into four layers. The record now covers three.
Controllers lost reliable communications at as many as 17 of 22 en route centers. None went dark.
Switches and routers overloaded with data and began giving inconsistent data, destabilizing voice and radar.
L3Harris copper-to-fiber work and a new Juniper router at El Dorado, Arkansas. Equipment at a radar site in Amarillo, Texas, installed weeks earlier, also flooded the network; the FAA blocked both sites.
The router or its installation, still undetermined; the maker says the device did not fail, and the FBI has joined the review. Why Amarillo joined in. And why two remote sites reached most of the centers and the backups at once.
Seven phases in about six hours.
From an unexplained halt to an official mechanism. The widest part of the vacuum came first: a national stop with no stated cause.
- 012:30 to 2:54 p.m.Local equipment outages
Kansas City Center and the Chicago airports report equipment problems. No national framing yet.
- 022:54 to 3:16 p.m.“Extraordinary circumstances”
The national stop arrives with no stated cause. The vacuum is at its widest.
- 033:16 to 4:05 p.m.A communications issue
The Transportation Secretary says all centers; the FAA says multiple. Anonymous FAA employees describe a rolling frequency outage.
- 044:05 to 4:30 p.m.A telecommunications outage
The FAA names the layer and says every flight is accounted for. Law enforcement’s preliminary read is no cyberattack; the DHS Secretary calls it “just a telecommunications outage as of now.”
- 054:30 to 5:30 p.m.The coincidence flood
Seven unrelated events are offered as connections, from a stale weather-radar page to the Iran conflict. The bar shows when the flood led the coverage this desk tracked. Narrative monitoring by NEATLABS PALISADE found the first link-seeking post about ten minutes after the stop, and a list joining several of these events before 4 p.m.
- 065:00 to 6:15 p.m.Competing cause threads
Bloomberg reports the FBI examining a software update; the FAA calls software claims false. CBS reports a network update crashed the L3Harris-operated network on reboot; a secondhand CNN account cites a router at Chicago Midway. Eurocontrol holds U.S.-bound departures.
- 077:16 p.m.Lift and mechanism
The stop is cancelled. At one White House briefing, two framings: the Transportation Secretary blames old routers and switches giving inconsistent data; the DHS Secretary describes patches and new technology overwhelming the older system.
L3Harris copper-to-fiber work is named, then the sites: El Dorado, Arkansas, and Amarillo, Texas. The router’s maker says it did not fail; the FBI joins the review; the FAA pauses new Juniper installations. Root cause analysis under way.
Times are Central, Friday 13 November 2026. Bars show when each framing led the coverage this desk tracked; phases overlap. Speculation in the wider conversation began earlier than these bars show.
Seven connections offered. None held.
While the cause was unnamed, at least seven unrelated events were offered as links to it. Two of them reached our own drafts before they were pulled.
A “National Radar Outage” page
An old, unmaintained NWS Anchorage page, quoted as current by national outlets. 158 of 159 weather radars were sending data.
An NWS network notice
The NWS did report nationwide internet and communications problems with its web products. Nothing ties it to the FAA.
The SMART rollout
An AI-supported planning tool reached all 22 en route centers that week. It carries no voice traffic; its developer denies a link. After the lift, the Transportation Secretary said the new AI-powered systems had nothing to do with it; the FAA Administrator, “zero evidence of anything AI-related.”
The Microsoft outage
Began about 20 hours earlier, with a stated non-hostile cause. DHS said the two were not connected.
A collision in New Orleans
A tanker struck a river cruise ship that afternoon. Nothing connects the two.
Drones over RAF Lakenheath
Added on a single posting time that put it twenty minutes before the stop, then withdrawn.
The Iran conflict
An active conflict made hostile intent easy to suggest. No evidence placed it in the event.
The one official move that shortened the vacuum was naming an open question early: the Homeland Security Secretary said investigators would examine the SMART rollout. A named question stops being a rumor.
Six snapshots of the live brief.
What the brief said at six points in the evening, from the status strip at the top of each update, and how each reading looks now.
- Ground stop
- Active
- Planned end
- 10 p.m. CT
- Failed layer
- Telecom, per FAA
- Ground stop
- Active
- Status
- Correction issued
- Failed layer
- Telecom, per FAA
- Ground stop
- Active
- Planned end
- Until further notice
- Failed layer
- Telecom, per FAA
- Ground stop
- Active
- Next review
- 0400Z, not a restart
- Failed layer
- Telecom; CBS: network update
- Ground stop
- Lifted, about 0116Z
- Cancellations
- 5,000+ Friday
- Failed layer
- Telecom; routers overloaded
Right early on the layer. Wrong seven times on the way.
The brief got the big calls right early: the failure sat in the ground network, the “frequency outage” described the ground side, the telecom and configuration hypotheses would merge, and cyber deserved restraint. Every change, and whether it held:
| Update | What changed | Held up? |
|---|---|---|
| Draft | National stop, rolling frequency outage, Chicago-area origin inferred. | Yes, including Chicago |
| 2 | Official communications description; scope points to a shared ground-network dependency; origin revised to unknown on a single-source Kansas City account. | Layer yes; origin revision no |
| 3 | “Severe but survivable,” inferred from airborne flights landing. | No; withdrawn in 4 |
| 4 | An unattributed report of failed backups; severity undetermined. | Yes; now more plausible |
| 5 | Telecommunications outage named; “accounted for” read as surveillance surviving. | Layer yes; surveillance reading no |
| 6 and 7 | SMART opened as a file; DHS confirms investigators will examine it. | Yes |
| 8 | Drafted, never sent: the weather service treated as a second agency failing; cyber raised. | No; stopped before it went out |
| 9 | Correction; the real weather-service notice discarded with the false one. | Partly; overcorrected |
| 10 and 11 | Software update contested; the two weather-service items separated. | Yes |
| 12 | Staged release; 0400Z as a review point, not a restart. | Yes; lifted earlier |
| 13 | L3Harris update account; Chicago and Dallas corrections. | Yes |
| 14 | An outside draft introduced four errors; fixed before it stood. | Yes after fixes |
| 15 | Eurocontrol hold; Lakenheath added, then pulled. | Yes |
| 16 and 17 | Lift; official mechanism; cyber closed; four-layer model; stale text cleared. | Yes |
The cyber rating moved four times. One move was an error.
The others followed real evidence, which is the discipline working as intended. The one driven by a mistake, Update 8, raised on a stale weather-service page, was stopped before it reached clients.
| Hypothesis | Path | Now |
|---|---|---|
| Shared telecom backbone | More likely from Update 2 onward; confirmed in layer by the FAA, in mechanism by the Transportation Secretary. | Mechanism confirmed |
| Software or configuration change | More likely; cut to Possible in 5; back to More likely in 13 as it merged with the first. | Trigger named Saturday: an equipment change to the shared network |
| Hostile cyber activity | Possible; Less likely (5); Possible (8, on a false signal); Less likely (9); Unlikely and closed (16). | Closed |
| RF jamming | Unlikely throughout. | Closed |
| SMART rollout | Opened in 6 as a timing question; examined officially; denied by its developer; dismissed by the Transportation Secretary after the lift. | Dismissed; FAA: zero evidence of anything AI-related |
Seven errors, three patterns.
We publish this because the patterns are general. They are the same three ways any team, newsroom or security desk goes wrong when a story outruns its facts.
Over-inference
- “Severe but survivable,” read from airborne flights landing (Update 3).
- “Surveillance survived,” read from “every flight accounted for” (Update 5).
- Lakenheath “twenty minutes before,” from a single posting time (Update 15).
Each read a precise conclusion out of a signal that could not carry it.
Premature discard
- The draft’s Chicago origin, dropped for a thinner Kansas City account (Update 2).
- The real weather-service network notice, discarded with the false radar page (Update 9).
- A 2025 Dallas outage precedent, doubted because it was unfamiliar.
Each threw out a true signal along with a weak or false one.
Provenance
- A stale NWS Anchorage page treated as an agency statement because national outlets quoted it (Update 8, stopped before it went out).
The vacuum’s output reached our own drafts within about half an hour.
Who was right, and how fast.
Official channels
Slow and partial. The FAA’s first substantive statement came about seventy minutes in, carried two of the four elements CB-060 calls for, and gave no next-update times. DHS carried more of the public explanation than the FAA did.
Unconfirmed reports
One unattributed report, of failed backups and aircraft without controller contact, grew more consistent with the official account: Nextgov later reported times when controllers could not immediately reach pilots or see aircraft. Another, of radars offline, echoed the stale public page. Weigh such reports by how the source could know.
Outside review
Strong as a lead generator: Eurocontrol, the software update, L3Harris and the Dallas precedent all surfaced there. Weak as an editor: one draft introduced four errors before it was corrected.
Major outlets
Generally reliable on what officials said. Weakest on provenance: national outlets amplified the stale radar page as current.
Where the story settled. What outlived the answer.
The official account took the lead within hours of the lift, but it did not clear the field. Frames that formed during the vacuum kept circulating after officials had answered them, and new ones grew in the space the root cause still leaves open.
The official technical account
Contractor fiber work at a remote site, a router flooding the network, legacy equipment giving inconsistent voice and radar data. No evidence of a cyberattack or of the AI tool.
Legacy infrastructure failure
Decades of deferred modernization, the 2009 router incident and repeated warnings, read as a system one routine change away from a national stop.
Contractor and funding incentives
Once the contractor was named, a frame took hold that the outage served the case for large modernization spending. It gained traction on X and little in mainstream coverage.
Cyber and sabotage
Claims of attack continued at lower volume after the FAA, DHS and the FBI said there was no evidence of one. Monday gave the frame fresh material: the Secretary called the router or its installation “the bad actor” and asked the FBI to join the review. Both are ordinary diligence; both read as a hint to anyone already looking for one.
Partisan blame
The President blamed the previous administration; replies disputed it with the record. Senate Democrats then pressed the administration on testing and backups. The frame moves the question from what failed to who is responsible.
Equipment and testing
With the router at the lab and installations paused, attention shifted from the contractor to the device and whether a sample-tested rollout reflected real operating conditions. This frame has a document trail: the senators’ letter and the pause itself.
The “tech wall”
On 16 November a software bug at Chase spread through shared payment rails and looked like a multi-bank outage. Chase said it was not cyber-related. Outlets joined it to the Microsoft outage and the ground stop as one story, and the clustering frame returned in mainstream form. The shape was the same: one failing node, many visible failures.
Coordinated disruption
The stop read as one of a set of same-day events. Officials treated them as unrelated, and nothing has connected them.
A few hours of vacuum produced frames that outlived the explanation. And the next stories formed where this page said they would: in the open root cause, once a contractor had a name.
How the two framings landed
Coverage leaned toward the Transportation Secretary’s account, old equipment and a funding emergency, over the Homeland Security Secretary’s new work overwhelming old systems. Some observers on X noticed the difference in emphasis as it happened.
The funding ask as fuel
The White House briefing after the lift also carried the Transportation Secretary’s $50 billion modernization request. The request helped seed the funding-incentive frame: official messaging feeding the space it was meant to close.
No operator required
NEATLABS PALISADE counted 84 posts from 55 independent voices across eight platforms and flagged none for coordination. The story kept moving to new platforms one to two days after the trigger was named, and no published fact-check matched its claims.
Two views of one event
This section draws on narrative monitoring by NEATLABS PALISADE from 13 to 15 November and on coverage through 17 November. The rest of the page is one desk’s live attribution record. The first shows what circulated and when; the second shows which stories had enough pull to bend a careful reader. Where they disagree, on when the flood began, the wider view was right.
It has happened before. Almost exactly.
Nationwide ground stops are rare: three in twenty-five years. Failures on the network behind this one are not, and the closest precedent ran through the same network, the same contractor and the same kind of fiber work.
| Nationwide stop | Cause | Duration and timing | Scale |
|---|---|---|---|
| 11 Sep 2001 | Security response to the attacks. All airborne aircraft ordered down. | About two days of closed airspace | All U.S. civil traffic grounded |
| 11 Jan 2023 | NOTAM database failure: contract staff deleted files while syncing the primary and backup databases. No cyberattack. | About 90 minutes, early morning, before the main departure bank | About 1,300 cancellations and more than 10,000 delays |
| 13 Nov 2026 | Router flood on the FAA’s contractor-run telecom network during copper-to-fiber work. No cyberattack. | About 4 hours 20 minutes, late afternoon, with aircraft and crews in motion | About 18,000 cancellations Friday to Sunday |
In November 2009 a technician for the same contractor misconfigured a router during a fiber transition. The error sat dormant for four weeks, until new work put another router in service, and then took down flight-plan systems nationwide.
The Department of Transportation’s Inspector General traced it in 2010. During Salt Lake City Center’s move to the fiber network, a Harris technician installed a temporary router at Los Angeles Center with a configuration error. Restoring the link masked it. On 19 November, a Harris engineer took the link down to bring a new router into service; the new router carried the same error, and the network failed across the country. Harris is now L3Harris, and the network is the same FTI.
In 2026, equipment installed at a radar site in Amarillo about three weeks earlier began flooding the network when the new router at El Dorado came online. Whether Amarillo carried a dormant error is the open question in Section 11. The 2009 record shows the mechanism exists on this network.
| Year | Where | What the backup did |
|---|---|---|
| 2004 | Los Angeles Center | The main voice system shut down; the backup crashed within a minute of activation. |
| 2007 | Jacksonville | All primary and alternate FTI services were lost. |
| 2009 | Salt Lake City and Los Angeles | A router failure spread nationwide; the FAA asked why the backup did not immediately resolve it. |
| 2023 | NOTAM system | The corrupted files were in the backup database too. |
| 2025 | Denver Center | Primary and main backup frequencies both failed; controllers fell back to an emergency frequency. |
| 2025 | Dallas | The FAA said contractor oversight failed to ensure the redundancies worked. |
| 2026 | 17 of 22 centers | Senators’ first question: why the backups did not contain the disruption. |
The backup shares the fault
For two decades the same pattern recurs: the redundancy rides the same network, holds the same data or carries the same configuration as the thing it backs up. The spread question on this page is a question about that pattern.
The same narrative arc
2009, 2023 and 2026 each ran from early speculation to a technical explanation to a call for modernization money. The arc is predictable enough to plan for: the funding ask arrives while the vacuum is still open.
Why 2026 was larger
Timing and duration more than severity. A ninety-minute morning stop in 2023 caught aircraft before the day began; a four-hour afternoon stop in 2026 stranded aircraft and crews mid-schedule, and recovery took the weekend.
Sources
DOT Inspector General report on the November 2009 FTI outage; FAA and contemporaneous reporting on 2004, 2007, 2023 and 2025; FAA questions and answers on the 2026 stop, 15 November.
The vacuum is narrower. It is not closed.
The trigger is on the record. The root cause and the spread are where the next stories will form: a trigger without an explanation of why it reached everywhere leaves the vacuum open.
- The root cause: the router or its installation. The device is at the L3Harris lab with FAA staff and FBI investigators; its maker says it did not fail; new Juniper installations are paused.
- The spread: why work at El Dorado set off equipment at Amarillo installed weeks earlier, and why two remote sites could destabilize as many as 17 of 22 centers and the backups at once. Senators asked the same question; a briefing is requested by 23 November.
- Whether the root cause analysis checks the Amarillo equipment for a dormant configuration error, the mechanism behind the 2009 outage on the same network.
- The remaining work: the switchover was part of the program replacing the network’s decades-old copper with fiber, which L3Harris said this month had reached 70 percent of FAA sites. How the rest is scheduled and tested is now a live risk; in September the GAO warned the FAA’s acceleration “could be prioritizing speed over costs and air traffic disruption.”
- Whether the FAA’s written account answers both, or names a trigger and leaves the spread unexplained.
- The doctrine: CB-312, on the attribution vacuum, carries this case and is open to all readers.
Interim analysis, updated 17 November 2026 with the router pause, the Senate letter, and narrative monitoring by NEATLABS PALISADE. The trigger is named; the root cause analysis is under way. This page will be revised when the FAA reports it.
Who is moving toward you, and what is taking shape.
That night needed both. One watches the principal; the other watches the story forming around the event.
- Starts with
- The principal, household or organization.
- On 13 November
- Movement plans, rebooking fraud lures, aircraft visible on public datalink feeds, exposure in crowded terminals.
- Starts with
- An issue, story or event. No name match needed.
- On 13 November
- The attribution vacuum, seven false connections, and the SMART story forming ahead of any evidence.