technology 6 min read

The FAA Routers That Grounded Flights Weren't Hacked—They Were Just

A nationwide flight shutdown traced to a configuration bug in aging L3Harris routers, not a cyberattack, exposes how every major aviation authority is tethered to a single vendor's obsolete hardware—and why that's a ticking clock.

  • Cybersecurity
  • Critical Infrastructure
  • Aviation Safety
  • FAA
  • Air Traffic Control

The router wasn’t broken. The system was.

The FAA’s nationwide ground stop that canceled more than 16,000 flights over two days began with something almost banal: an engineer turning on a new router at a facility in El Dorado, Arkansas. The device—a Juniper Networks unit installed weeks earlier by contractor L3Harris Technologies—broadcast routing data across the entire network instead of its local segment. The broadcast storm hit a second, similarly configured router in Amarillo, Texas, which responded by flooding the system further. Within hours, the FAA’s communications backbone was saturated. Planes sat on tarmacs. Controllers lost contact. The sky went quiet.

Officials confirmed Sunday there was no sign of a cyberattack. What they did confirm is uglier: the U.S. air traffic network is held together by aging commercial-grade routers running a feature that was never supposed to run at this scale, inside a system a 2023 FAA-commissioned safety review called obsolete.

This isn’t a cybersecurity story. It’s a dependency story.

One vendor holds the keys to the skies

L3Harris owns and operates the FAA’s en route automation system—the network that passes voice and data between controllers and aircraft across the continental United States. The contract has kept the company as the sole provider for more than a decade. Every upgrade, every router swap, every software patch flows through a single procurement line. That concentration of responsibility is not unusual in government contracting, but in aviation it carries consequences most policymakers don’t discuss publicly.

When the Arkansas fiber upgrade triggered the cascade, the response wasn’t to reroute around a failed node. There is no alternative path. The FAA turned off the broadcast feature across all 69 routers of that type in the network, a bandage that restored function but did nothing to address the underlying architecture. Transportation Secretary Sean Duffy said Sunday the department is still weighing whether the incident resulted from “incompetence” or faulty hardware. Both answers point to the same structural weakness: one vendor, one design, one point of catastrophic failure.

The history repeats. In 2009, a technician for Harris Corporation—the company L3Harris acquired—misconfigured a router during a similar copper-to-fiber migration, disrupting 129 air traffic facilities. The same mistake, fifteen years apart, different contractor, identical root cause.

What happens next depends on who funds the fix

President Trump used the outage to reignite a familiar demand: $50 billion for a complete replacement of the en route automation system, which he calls FAST. The previous administration approved $12.5 billion last year. Whether Congress will appropriate another $37.5 billion on top of that while airlines are still canceling flights remains an open question.

Duffy cautioned that more disruptions are likely before the copper-to-fiber transition finishes at the end of 2028. That timeline spans two full congressional sessions and whatever budget negotiations follow. Each delay keeps the current router fleet in service—one model, one vendor, one broadcast feature that now exists in every node.

Travelers are left navigating the aftermath alone. American Airlines extended its travel waiver through Wednesday. JetBlue canceled roughly a quarter of its schedule on Sunday. Refunds are owed for canceled flights, but Hotel and meal rebooking under DOT rules generally covers only disruptions within an airline’s control—an FAA ground stop doesn’t qualify. The financial pain lands on passengers; the political accountability drowns in process.

The global implication nobody is tracking

The FAA’s predicament isn’t unique to the United States. Across Europe, Asia, and the Middle East, air navigation service providers depend on a shrinking pool of avionics manufacturers. The three companies that dominated this space thirty years ago have consolidated into roughly five, and each provider has built its operational network around a single supplier’s equipment. When Juniper told the Wall Street Journal its devices “did not fail,” it was making a narrow technical claim—but the broader reality is that no alternative device was evaluated for this deployment. The broadcaster feature existed because a single vendor designed the protocol, tested it in limited conditions, and shipped it to a monopoly customer.

Other nations face the same architecture with their own vendors. Eurocontrol’s members operate networks where hardware sourcing is similarly constrained. Australia’s Airservices, Canada’s NAV CANADA, and Japan’s Mitsui OSK Lines-affiliated navigation authority all manage aging infrastructure upgrading onto fiber, all under the same vendor consolidation pressures. None of them publish their router counts or configuration standards in real time.

The difference between a broadcast storm and a successful attack can be thin in practice. The FAA’s outage demonstrated that the system could be brought down without any malicious actor entering it. An adversary who understands the network topology and the router firmware would need far less effort than most assume. This isn’t fearmongering—it’s arithmetic.

The numbers don’t lie

The trajectory this week tells the story faster than any policy brief:

  • Friday: 9,000+ cancellations
  • Saturday: 6,700 cancellations
  • Sunday morning: 1,570 cancellations, still above normal but declining
  • Estimated return to baseline: end of Sunday

American and JetBlue absorbed the steepest share of Sunday’s remaining cuts. FlightAware tracked the recovery. But the cancellation count is a lagging indicator. The actual disruption includes crews stranded hundreds of miles from home, aircraft positioned incorrectly across the country, and schedule fragments that won’t resolve until Monday’s morning rush absorbs them. Airlines typically need 48 to 72 hours after a major ground stop to fully restack their operations.

Bryan Bedford, the FAA administrator, says the agency expects to be back to normal by the end of Sunday. That prediction may prove accurate for the booking system. Recovery for the physical network—planes, people, slots—is slower and less visible.

What actually needs to change

The immediate fixes are straightforward: keep the broadcast feature disabled, accelerate the remaining copper-to-fiber conversions at sites that haven’t completed the swap, and audit every node in the 69-router fleet for the same misconfiguration pattern. None of these require billions.

The structural fix is harder. The FAA needs a procurement framework that treats network redundancy the way it treats aircraft redundancy—no single point of failure permitted on the flight deck, and none should be acceptable on the ground either. That means either multi-vendor architectures or contractual guarantees that force a single supplier to maintain competitive pricing and hardware diversity. Both approaches conflict with how federal acquisition currently works, which is precisely why nothing has changed since 2009.

Duffy called the engagement from the president “remarkable.” Remarkable engagement without appropriations is just theater. The $50 billion FAST request will survive a committee hearing and emerge diminished—or it won’t surface at all. Either outcome leaves the current fleet in place through 2028, running a feature everyone now knows is dangerous, inside a network nobody can reroute around.

The router wasn’t hacked. That should make every aviation authority uncomfortable, not relieved.