Pilot warned Amazon jet was going too fast

Gammatek ISPL

Sep 10

11 min read

Author: Gammatek ISPL, Published: Sep 2026

Nearly two minutes before an Amazon-branded cargo jet slammed through a perimeter fence and killed five people near Miami International Airport, one of the two pilots in the cockpit repeatedly said the plane was coming in too fast and called for a go-around. According to the National Transportation Safety Board, the other pilot did not consistently respond. That gap, between a warning clearly spoken and a decision not taken, is the actual story here, and it's one that applies far beyond aviation. Any organization that relies on a junior person to challenge a senior one in the middle of a fast-moving, high-consequence decision has some version of this same failure mode sitting inside its operations right now, waiting for the wrong day to show up.

What happened, according to the NTSB

On September 6, 2026, Amazon Air Flight 7598, a Boeing 767-300 operated by North Carolina-based cargo carrier 21 Air on Amazon's behalf, was completing its third flight of the day, a short hop from San Juan, Puerto Rico back to Miami, when it came in for landing on runway 30 at Miami International Airport. According to NTSB Chair Jennifer Homendy, the aircraft overran the runway by roughly 1,300 feet, struck airport navigational equipment, tore through the perimeter fence, and collided with two vehicles on the road beyond it, a van carrying aircraft-cleaning crew members and a passing SUV. Five people in those vehicles were killed: Rolando Aleman Leon, 55; Yoel Rodriguez Narajo, 53; Julio C. Pineda, 75; Carlos Acosta Fajardo, 53; and Javierkys Reyes Quevedo, 47. Five more people were injured, three of whom remained hospitalized as of the NTSB's briefing. Both pilots survived and were released from the hospital.

The NTSB recovered both the flight data recorder and the cockpit voice recorder, and its preliminary review of that cockpit audio is what produced the finding driving national coverage: one of the two pilots repeatedly warned, starting nearly two minutes before landing, that the aircraft was approaching too fast, and urged a go-around, essentially aborting the landing and circling back for another attempt. The NTSB said the other pilot did not consistently respond to those warnings. A former FAA official, Mike O'Donnell, told reporters that two minutes is a substantial window in cockpit terms, plenty of time to slow the aircraft or simply commit to a go-around, and that the plane's nose-and-right-gear-first touchdown is itself evidence of an unstable approach, the exact condition that go-around procedures exist to catch before it becomes unrecoverable. The full investigation, including a determination of probable cause, is expected to take a year or more.

This is also the second fatal cargo jet accident in the United States in under a year, following a UPS cargo aircraft crash at Louisville's Muhammad Ali International Airport the previous October. Two serious cargo-aviation accidents in quick succession is the kind of pattern that tends to trigger closer regulatory scrutiny of cargo-carrier safety programs generally, independent of what each individual investigation ultimately finds.

Why cargo carriers see this pattern more than passenger airlines

There's a structural reason cargo operations carry a different risk profile than passenger aviation, and it's worth naming explicitly rather than treating this crash as an isolated anomaly. Cargo aircraft in Amazon's contracted network, and across the industry generally, are frequently older airframes bought after they've aged out of passenger service. The aircraft involved in the Miami crash was 32 years old, originally built and flown as a passenger jet for multiple airlines before being converted to a freighter configuration in 2015, a common and FAA-regulated path for aircraft nearing the end of their passenger-carrying life but still airworthy for cargo use under a different maintenance and inspection regime. Age alone doesn't cause an accident; commercial aircraft routinely fly safely for well over 30 years under proper maintenance. But it does mean cargo fleets, as a category, skew toward older equipment with longer maintenance histories to track, which is one more variable investigators typically examine alongside crew performance and procedures.

The scheduling pattern is also worth noting without over-reading it. Flight 7598 was on its third leg of the day at the time of the crash, having already flown Cincinnati to Miami and Miami to San Juan before the final return leg into Miami. Multi-leg cargo rotations are standard in the industry and are governed by the same FAA duty-time and rest-requirement rules that apply to passenger crews. Whether fatigue played any role in this specific crew's performance is a question for the NTSB's ongoing review, not something the preliminary facts currently establish either way, but it's the kind of factor a thorough investigation routinely rules in or out before reaching a final determination.

None of this is offered to assign blame ahead of the NTSB's findings. It's offered because any business operating its own vehicle fleet, whether that's delivery vans, cargo trucks, or a chartered aircraft, inherits some version of the same variables: aging equipment, multi-leg or multi-stop schedules, and crew fatigue, all layered on top of whatever human-communication failure ultimately gets identified as the proximate cause. A safety review that only asks "did someone speak up" without also asking "was the equipment, the schedule, and the rest cycle set up to make speaking up more or less likely to matter" is examining half the system.

Why "he warned them and they still crashed" is the real finding

It would be a different story, and a simpler one, if nobody in that cockpit had noticed the aircraft was coming in hot. That's not what the preliminary evidence shows. Somebody noticed. Somebody said something. Repeatedly. Nearly two minutes out, which by the account of an aviation safety veteran was enough time to do something about it. And the flight continued to landing anyway.

That specific pattern, a valid warning correctly identified and then not acted on, is a known and heavily studied failure mode in aviation safety research, and it's not primarily a story about one distracted or overconfident pilot. It's a story about what happens inside a two-person hierarchy under time pressure when one person's warning doesn't translate into the other person's action. Whether that came down to the warning not being forceful enough, not being heard clearly enough, or being heard and set aside, is exactly what the NTSB's ongoing review of the cockpit voice recorder is meant to establish. But the shape of the failure, a warning that didn't convert into a decision, is one aviation has a specific vocabulary for, and it's worth understanding regardless of how this particular investigation concludes.

The authority gradient: the decades-old aviation research this maps onto

Aviation safety researchers have a specific term for the dynamic where a valid safety concern, raised by a less senior or less dominant crew member, fails to produce action from the person actually flying the aircraft: the authority gradient. It refers to the perceived or actual difference in power, status, or seniority between two people making a decision together, and how that difference shapes whether the person with less authority actually pushes a concern hard enough to change the outcome, rather than raising it once and then deferring.

The concept isn't new, and it isn't theoretical. It traces back to the 1977 Tenerife disaster, still the deadliest accident in aviation history, where a steep authority gradient in the cockpit is widely cited as a contributing factor in the captain's decision to take off without clearance, a decision his first officer sensed was wrong but didn't forcefully contest. It shows up again in a well-documented 2009 icing-related crash, where the NTSB's final report specifically found that the first officer, despite her own concerns, deferred to the captain's decision not to go around, later telling investigators she assumed he "had a good reason" and trusted his judgment even as her own instincts said otherwise. The NTSB's response to that crash was a formal recommendation that crew resource management training explicitly include assertiveness training for the less senior pilot, specifically because a warning raised once and then dropped doesn't function as a warning in any way that changes what happens next.

The pattern across decades of these incidents isn't that co-pilots don't notice problems. It's that noticing a problem and successfully converting that observation into a changed decision are two different skills, and the second one is far harder to train, especially under time pressure, especially across a real difference in seniority, and especially when the senior person isn't visibly receptive to being challenged in the moment.

Incident	Year	What the warning looked like	What happened to the warning
Tenerife (KLM/Pan Am collision)	1977	First officer sensed the takeoff clearance wasn't fully confirmed	Deferred to the captain's seniority; did not forcefully press the concern
Colgan Air 3407 (icing-related crash)	2009	First officer had concerns about continuing the approach in icing conditions	Assumed the captain "had a good reason," did not raise it again
Amazon Air Flight 7598 (Miami)	2026	Pilot repeatedly warned the aircraft was approaching too fast, nearly two minutes before landing	NTSB says the other pilot did not consistently respond; investigation into why is ongoing

None of these three incidents are the same in their specifics, and it would be premature to draw a final conclusion about Flight 7598 before the NTSB completes its work. But the shape connecting them, across five decades of aviation safety research, is worth sitting with regardless of how this specific investigation resolves: the hard part of safety culture was never getting someone to notice a problem. It's building a system where noticing a problem reliably produces a changed decision, even when the person noticing it is not the one with final authority.

Comparable incidents, and what changed after them

Aviation's response to authority-gradient failures over the past several decades hasn't been to hope pilots individually get better at speaking up. It's been structural: standardized callouts that require a specific verbal response rather than a nod, sterile-cockpit rules that reduce distraction during critical phases of flight, and formal unstable-approach criteria that convert "something feels off" into an objective checklist, if the aircraft exceeds specific speed, altitude, or configuration parameters below 1,000 feet, a go-around isn't a judgment call anymore, it's the mandatory procedure. The entire point of that shift was to stop relying on a junior person successfully overriding a senior person's momentum in the moment, and instead build a rule that removes the need for that confrontation altogether.

That shift is also, not coincidentally, the same one crew resource management researchers have spent years trying to export into healthcare, particularly intensive care units, where the same hierarchy-driven silence has been directly linked to preventable patient harm. The lesson exported from aviation to medicine wasn't "train people to be braver." It was "build a system where a specific, trained phrase or action triggers a mandatory response, so the outcome doesn't depend on how brave any one person feels able to be that day."

Implementation considerations for any high-stakes team, not just cockpits

You don't need a cockpit for an authority gradient to cost you money, safety, or reputation. Warehouse floor supervisors who don't feel able to stop a shipment a more senior manager wants pushed out the door. A junior security analyst who flags a suspicious login and gets waved off by a senior engineer who assumes it's a false positive. A newer technician who notices a fleet vehicle failed a pre-trip inspection item but doesn't feel empowered to hold up the route. The structural fix aviation landed on after fifty years of studying this problem applies just as directly to any of these situations.

Replace "raise a concern" with a mandatory callout-and-response protocol for your highest-risk decisions. If a specific role notices a specific defined condition (a failed inspection item, an anomalous system alert, a shipment that doesn't match its manifest), require a documented response from the decision-maker, not just an opportunity to mention it.
Set objective, numeric thresholds wherever you can, instead of relying on judgment calls under time pressure. Aviation's unstable-approach criteria work precisely because they remove ambiguity: hit the threshold, execute the mandatory action, no debate required in the moment.
Train your senior people specifically on how to receive a challenge, not just your junior people on how to raise one. Most safety training over-invests in teaching juniors to speak up and under-invests in teaching seniors to visibly, immediately take a raised concern seriously, which is the half of the exchange that actually determines whether the next junior person bothers speaking up again.
Audit near-misses for silence, not just for errors. A near-miss report that says "no one flagged it" is a different and often more urgent finding than one that says "someone flagged it and the fix worked." A near-miss where someone flagged it and the outcome was still bad, the way this preliminary NTSB finding reads, is the most urgent finding of all.
Review whether your incident response and safety-reporting systems can actually reconstruct who said what and when. The entire reason the NTSB can say a warning happened nearly two minutes before impact is because the cockpit voice recorder existed and was recovered. Most non-aviation organizations have no equivalent record of the warning that got raised and ignored last quarter, which means the pattern repeats invisibly until it doesn't.
Separate the equipment and scheduling variables from the communication variable in your own audits. As this incident illustrates, a communication failure rarely happens in isolation from the conditions around it, aircraft age, route fatigue, time pressure. A safety review that only interviews staff about "did you feel able to speak up" without also auditing the equipment condition and schedule pressure they were operating under will miss half the causal picture.

Frequently asked questions

What actually caused the Miami cargo jet crash? That determination is still ongoing. The NTSB's preliminary review found that one pilot repeatedly warned the aircraft was approaching too fast and called for a go-around nearly two minutes before landing, and that the other pilot did not consistently respond. The full investigation into why, including crew, aircraft, and procedural factors, is expected to take a year or more before a final probable cause is issued.

Was the age of the aircraft a factor in the crash? It hasn't been established as a cause. The aircraft was 32 years old and had been converted from passenger to cargo configuration in 2015, which is a common and regulated practice in the cargo industry. Investigators typically examine aircraft maintenance and mechanical history alongside crew performance, but no findings linking the aircraft's age directly to this crash have been released as of the NTSB's preliminary briefing.

What is an "authority gradient" in aviation? It's the difference in perceived authority, seniority, or status between two people making a decision together, most often a captain and first officer, and how that difference affects whether a raised safety concern actually changes what happens next. A gradient that's too steep can make a junior crew member reluctant to press a valid concern past a single mention.

Is this the same incident as the 2019 Amazon Prime Air crash? No. The 2019 crash, Atlas Air Flight 3591, involved a different aircraft and a different operator, and killed all three people aboard after a loss of control in flight near Houston. The two events share only the Amazon-branded cargo service; the aircraft, operators, and circumstances are unrelated.

Does Amazon operate its own cargo planes? No. Amazon Air's flights are operated by contracted cargo carriers rather than by Amazon directly. Flight 7598 was operated by 21 Air LLC, a Miami-based cargo carrier, on Amazon's behalf.

How does this apply to a business that isn't in aviation? Any team where a less senior person needs to be able to stop or redirect a decision made by a more senior person, in a fast-moving, high-consequence situation, has some version of an authority gradient. Warehouse operations, fleet management, security operations, and healthcare all have documented versions of the same failure pattern, and the structural fixes aviation developed (mandatory callouts, objective thresholds, training seniors to receive challenges) transfer directly.

A safety and compliance review built for exactly this failure pattern

If your organization runs safety-critical operations, fleet or logistics management, warehouse floors, security monitoring, and you're not confident a junior employee's flagged concern would actually change a senior decision-maker's next move in the moment, that gap is worth finding before an incident forces the question. Our operational safety and compliance review maps exactly where your escalation paths depend on individual courage rather than mandatory protocol, and helps you build the structural fix instead.

https://www.gammateksolutions.com/post/amazon-prime-members-just-got-a-huge-new-perk-here-s-how-to-use-it https://www.gammateksolutions.com/post/the-best-worst-and-strangest-ways-ai-is-really-being-used-at-work https://www.gammateksolutions.com/post/big-tech-profits-get-160bn-boost-from-gains-on-stakes-in-other-ai-companies