Daniel sent us this one — and it's one of those questions that sounds simple until you actually try to answer it. Do air marshals really exist? How many flights are they actually on? And do the cabin crew know who they are? The short answers are yes, about one point four percent of flights, and it's complicated in ways that get genuinely strange. But the real story here is the gap between what Hollywood tells us and what's actually happening in the skies. Welcome to My Weird Prompts, episode two hundred one.
The air marshal is the Loch Ness Monster of aviation security. Everyone's heard of them, a few people claim to have seen one, and the government insists they're real while keeping almost everything about them classified. And the reason this matters right now — I mean, beyond it just being a fascinating piece of bureaucratic theater — is that the TSA is facing a pretty uncomfortable congressional moment. There was a GAO report, originally from twenty twenty-five, that got leaked through a FOIA request back in March. The headline number: FAMS operational costs have gone up eighteen percent since twenty twenty-three, while the number of disruptive passenger incidents where they actually intervened dropped by twelve percent.
We're spending more to get less. That's the congressional math problem.
And the report — the designation is GAO dash twenty-five dash one zero six four seven eight — it's the kind of document that makes budget committee staffers reach for the antacids. The FAMS budget for fiscal year twenty twenty-five was eight hundred and five million dollars. For a program that covers, by the GAO's own numbers, one point four percent of domestic flights.
One point four percent. So if you're on a plane right now, statistically, there's about a one in seventy chance a marshal is on board. Those are lottery ticket odds.
We haven't even gotten to the weird part yet. The weird part is who knows about them, how they get assigned to flights, and what they're actually legally empowered to do. Which turns out to be very different from what most people assume.
Alright, let's start with the basics. Do air marshals actually exist? The short answer is yes, but the long answer is where it gets weird.
The Federal Air Marshal Service is a component of the TSA. It was created in its current form after nine eleven, but it's actually built on top of an older program. The original Sky Marshal program started in nineteen sixty-one under Kennedy, and it was a direct response to what they called the Cuban hijackings. Planes getting taken to Havana was basically a recurring news item. The program was small, it was run by the FAA, and it kind of faded in and out of relevance depending on the threat landscape.
The hijackings to Cuba thing is worth pausing on, because it sounds almost quaint now. Someone takes over a plane, demands to go to Havana, and everyone just sort of grumbles about the delay. It wasn't the kind of existential threat we think about today.
Right, and that shift in the nature of the threat is central to understanding the modern FAMS. After nine eleven, the nine eleven Commission Report — specifically recommendation three point two — mandated a massive expansion of the program. The logic was that the old model of hijacking, where you negotiate and eventually land somewhere, had been replaced by the suicide hijacking model. There's no negotiating with someone who doesn't intend to land. So you need armed intervention capability on the plane itself. The question that's been hanging over the program ever since is whether that capability is actually being deployed in a way that matches the threat.
The numbers tell a story that's hard to square with the public perception. Walk me through the math.
To understand how they work, we have to look at the numbers. And the numbers are surprisingly small. The FAMS has somewhere between three thousand and four thousand active agents. The exact number is classified, but based on budget documents and the leaked GAO report, that's the range most analysts point to. Now, on any given day in the United States, there are roughly twenty-five thousand to thirty thousand domestic flights. If every single marshal flew every single day — no training days, no admin time, no mandatory rest periods — you'd cover at most about fifteen percent of flights. But they don't fly every day. Former agents, in interviews with The Verge back in twenty twenty-four, estimated the actual coverage is closer to one to two percent of daily flights. The GAO report confirmed that: one point four percent in twenty twenty-four.
The image of the stoic guy in the suit, sitting in seat three C, scanning the cabin — that's basically a myth for ninety-eight point six percent of flights.
It's a myth that the TSA is perfectly happy to let persist. And this is where we get into what I think is the most intellectually interesting part of the whole program. The concept is called randomized deterrence. The TSA deliberately leaks the existence of the program. They want potential attackers to believe a marshal might be on any flight. The cost of the program — eight hundred and five million dollars a year — is justified not by the number of arrests made, but by the number of attacks that never happen. Which is, by definition, a non-falsifiable claim.
You can't prove a negative. So the program's success metric is something you can never actually measure.
And this is the kind of thing that drives budget analysts absolutely insane. The TSA can point to the absence of a successful in-flight terrorist attack on a US carrier since nine eleven and say, see, it's working. A critic can say, well, it's working because the cockpit doors are reinforced and passengers now know to fight back, and you have no way of knowing whether the air marshals contributed anything. Both positions are logically coherent and neither can be disproven.
It's the aviation security equivalent of the elephant repellent argument. I don't see any elephants around, so the repellent must be working.
That is a shockingly apt analogy. And it gets even more interesting when you look at how they decide which flights actually get a marshal. They're not on random flights. They're deployed based on a daily threat matrix. This matrix combines intelligence from the FBI's Joint Terrorism Task Forces, the TSA's No Fly and Selectee lists, and something called CAPPS — the Computer Assisted Passenger Prescreening System. The current version is a derivative of CAPPS two and three, and the algorithm itself is classified as Sensitive Security Information, or SSI.
We don't know exactly what factors it weighs.
We know the broad strokes. It looks at origin and destination cities, the passenger nationality mix on a given flight, recent threat intelligence from the intelligence community, and the individual risk scores of passengers who've been flagged by the system. If a flight has a higher concentration of passengers with elevated CAPPS scores, it's more likely to get a marshal. The problem, and this is something critics have been pointing out for years, is that this system is designed to catch networked threats — people connected to known terrorist organizations, people on watchlists, people whose travel patterns match a known profile. It's largely useless against a lone wolf attacker who has no prior intelligence footprint.
Which is increasingly the threat model we're worried about. The guy who radicalizes in his basement, buys a ticket with a credit card, and shows up at the airport looking like everyone else.
If that person doesn't trigger any of the CAPPS flags, the algorithm has no way of flagging their flight as high-risk. So the marshal goes to a different flight — maybe one with a Saudi Arabian exchange student who's on a watchlist because his cousin once attended a mosque that was under surveillance. The algorithm is doing exactly what it was designed to do, but what it was designed to do may not match the actual threat landscape.
There's a case study that illustrates this whole thing, right? The shoe bomber copycat incident.
In twenty fifteen, on a Delta flight from Atlanta, a passenger attempted to ignite a device in his shoe. It was a direct copycat of Richard Reid from two thousand one. A federal air marshal was on board and subdued the passenger. This is, to my knowledge, the only publicly confirmed FAMS intervention on a US domestic flight in the last decade. Every other in-flight security incident you've heard about — the passengers who tried to open cockpit doors, the fights over masks, the drunk guys who had to be duct-taped to their seats — those were handled by flight attendants and fellow passengers.
The one confirmed case where a marshal actually did the thing they're trained to do was ten years ago.
I should note, there may be other interventions that were never made public. The FAMS is extremely tight-lipped about operational details. But the fact that we can only point to one in a decade is telling. Compare that to the UK, by the way. Their equivalent is SO15, which is part of the Metropolitan Police. It operates on a much smaller scale, primarily on high-risk international routes to and from the Middle East. The British model is more targeted, more transparent about its limitations, and doesn't carry the same kind of Hollywood mythology that the American program does.
Then there's Israel, which is a completely different animal.
Israeli air marshals — they're Shin Bet, not a separate agency — are always armed with automatic weapons, they travel in teams of two or three, and the crew knows exactly who they are. It's a fundamentally different operational doctrine based on a fundamentally different threat model. Israel faces a higher volume of specific, credible threats against its aviation sector, and it responds with a more aggressive, less secretive posture. The American model prioritizes secrecy and randomized deterrence. The Israeli model prioritizes overwhelming force and crew coordination.
Which brings us to the question of who actually knows the marshal is on board. Because that's where this gets strange.
We know they exist, and we know they're rare. But the most fascinating part is how they interact with the crew and the legal system. The official policy — and this comes from the FAMS Standard Operating Procedure, which was leaked through MuckRock in twenty twenty-two — is that the captain is informed of the marshal's presence during the pre-flight briefing. They use a secure code word.
What's the code word?
Of course it is. It's like they let a twelve-year-old name the protocol.
I cannot improve on that observation. It's Cobra. The captain is told, a marshal is on board, here's their seat number, here's the code word. The flight attendants are not told. They are explicitly kept in the dark. The marshal may reveal themselves to the lead flight attendant in an emergency, but doing so casually is a breach of protocol.
You've got a cabin full of flight attendants who are supposed to manage passenger safety, and there might be a federal agent sitting among the passengers, and they have no idea who it is or even if they exist on this particular flight.
The logic behind this is what they call the insider threat. If a flight attendant is compromised or coerced — say someone threatens their family — they can't reveal the marshal's identity if they don't know it. The marshal's anonymity is their primary tactical advantage. The problem, and this is something pilot unions have been vocal about, is that this also means there's no coordination mechanism in an emergency. The marshal acts independently. If a situation escalates, the crew doesn't know who's about to stand up and draw a weapon. The marshal doesn't know what the crew is doing to manage the situation from their end.
You've got two separate security responses running in parallel with no communication between them.
That's not a bug, from the FAMS perspective. That's the feature. They want the marshal to be able to assess the situation without being influenced by the crew's assessment, and they want the crew to continue their normal emergency protocols without relying on the marshal's presence. But in practice, it creates exactly the kind of friction you'd expect. There was an incident in twenty twenty-one on a Southwest flight — a passenger tried to open the cockpit door. The marshal on board was seated in row twenty-eight. By the time he fought his way forward through the aisle, passengers had already subdued the attacker. The marshal's report was classified, but the fundamental problem is obvious: if your intervention force is stuck in a middle seat at the back of the plane, your response time is measured in minutes, not seconds.
Which gets to something I hadn't even considered. Marshals are subject to the same seat assignment lottery as everyone else?
This is the deadhead problem, and it's one of those operational realities that sounds absurd but makes perfect sense once you think about it. Marshals often fly in plain clothes, posing as regular passengers. They can't demand a specific seat — like an exit row or a bulkhead — without blowing their cover. If they walk up to the gate agent and say, I need seat three A for operational reasons, they've just announced to everyone at the gate that they're a federal agent. So they get assigned whatever seat the algorithm gives them. Middle seat, back row, next to the lavatory — they have to take it and make it work.
The image of a federal air marshal, armed, trained, ready to neutralize threats, crammed into seat twenty-eight B with his knees up against the tray table, is the perfect metaphor for this entire program.
It really is. And it raises a genuine tactical question. If a threat emerges in first class and your marshal is in the last row of economy, what exactly is the response plan? Fight through a narrow aisle full of panicked passengers? That's not a plan, that's a hope.
Let's talk about what they're actually legally empowered to do, because I think most people assume they're counterterrorism operatives, and the reality is more complicated.
Air marshals are federal law enforcement officers under forty-nine USC section forty-four nine oh three. They have full arrest powers for federal crimes committed on aircraft. But here's the thing: their primary day-to-day function, in terms of what they actually do that generates paperwork, is handling unruly passengers. Drunk people, violent people, people who refuse to comply with crew instructions. They make arrests that flight attendants legally cannot make.
They're less Jason Bourne and more the police officer you call when a guy at a bar won't stop bothering people.
That's the mission creep issue. There was a high-profile incident in twenty twenty-three where a marshal arrested a passenger for refusing to wear a mask. This sparked a significant debate within the agency and in the press about whether FAMS resources should be used for what amounts to a compliance enforcement role. The counterargument from the TSA is that any disruptive passenger could be a distraction tactic for a more serious attack, so all disruptions are security-relevant. But that's the kind of logic that justifies literally any use of the program for literally any purpose.
It's the TSA version of qualified immunity. If everything is potentially terrorism, then nothing is out of scope.
This connects to the training question, which is another area where public perception diverges from reality. The image of the air marshal as a special forces operative — someone with SEAL Team Six level training who just happens to be sitting in coach — that's a Hollywood invention. Air marshals are TSA employees who undergo a twelve-week training course at FLETC, the Federal Law Enforcement Training Center. That's less rigorous than FBI training, less rigorous than Secret Service training. It's a solid law enforcement training program, but it's not a tier one special operations pipeline.
So you could go from working at a TSA checkpoint to being a federal air marshal in about three months.
The hiring pipeline is more selective than that — most marshals have prior law enforcement or military experience — but the training itself is twelve weeks. And the curriculum covers firearms, defensive tactics, close-quarters combat in an aircraft environment, and legal authorities. It does not produce Jason Bourne.
The other thing I want to dig into is the psychological dimension of all this. Because the marshal's job, on the vast majority of flights where nothing happens, is to sit in a seat for hours, pretending to be a regular passenger, while maintaining constant situational awareness. That's an incredibly cognitively demanding task with essentially zero stimulation.
The burnout rate in FAMS is notoriously high, and this is a big part of why. Imagine your job is to watch people read magazines and eat pretzels for four hours while remaining at a heightened state of alertness, knowing that the one time you slip into autopilot could be the one time something happens. It's a recipe for chronic stress. There have been internal reports — not public, but referenced in GAO testimony — about high rates of alcohol abuse, marital problems, and mental health issues among marshals. The job eats people.
Sitting alone with your thoughts for hours while armed and on alert. That's not a job, that's a psychological experiment.
The isolation compounds it. They can't tell their families where they're going. They can't maintain friendships with colleagues in a normal way because operational schedules are compartmentalized. They're essentially undercover for their entire career, but without the adrenaline payoff that undercover work in other agencies provides. Most undercover FBI agents eventually get the raid, the arrest, the resolution. A marshal might spend twenty years on planes and never draw their weapon outside of the firing range.
You've got a workforce that's burning out, a deployment algorithm that covers less than two percent of flights, a secrecy protocol that prevents coordination with the crew, and a mission that has shifted from counterterrorism to handling drunk passengers. And we're spending eight hundred and five million dollars a year on this.
Which brings us to the question of whether this program is actually effective against modern threats. And I want to be fair here — there are smart people who believe the deterrence value justifies the cost. The argument is that even if the marshal never fires a shot, the uncertainty they create forces potential attackers to plan around an unknown variable. That makes attack planning harder, which makes attacks less likely. The problem is that this logic works best against the kind of organized, networked terrorist groups that were the primary concern in two thousand one. It works much less well against the kind of threat we're seeing more of now — the lone actor, the disturbed individual, the person who isn't doing careful operational planning.
The guy who decides on Tuesday that he's going to do something on Thursday doesn't care about the randomized deterrence model because he's not running a cost-benefit analysis.
And this is where the CAPPS system really shows its age. It was designed in a post-nine eleven world where the primary threat was Al Qaeda-style terrorist networks. It looks for patterns, connections, travel histories, financial trails. It's a network analysis tool. A lone wolf with a clean record and a one-way ticket purchased yesterday doesn't trigger any of those flags. And if the algorithm doesn't flag the flight, the marshal goes somewhere else.
All of this raises a practical question for the average flyer. What does this mean for you on your next flight?
The first thing to understand is that the security theater you experience at the airport — the pat-downs, the shoe removal, the liquid restrictions — that's the visible deterrent. The air marshal program is the invisible deterrent. And statistically, it's almost certainly not on your flight. Your safety, in the extremely unlikely event of an in-flight security incident, relies almost entirely on the flight crew and your fellow passengers. That's not speculation. That's what the data shows. The vast majority of in-flight interventions are performed by flight attendants and passengers, not marshals.
If you see something, the protocol is not to look around for the guy in the suit who might be a federal agent. The protocol is to alert a flight attendant immediately.
And here's the thing about that protocol: if a marshal is on board, they will likely identify themselves only after the situation escalates beyond the crew's control. The marshal's job is to be the last resort, not the first responder. The first responders are the flight attendants. They're trained in de-escalation, they have restraints on board, and they have direct communication with the cockpit. The marshal is sitting in seat fourteen C waiting to see if this is a real threat or just a guy who had too many airport Bloody Marys.
The marshal can't communicate with the crew to make that assessment, because the crew doesn't know who they are.
Which means the marshal is making a unilateral judgment call about when to break cover, based on incomplete information, while the crew is simultaneously managing the situation based on their own incomplete information. It's a system that works despite itself, when it works at all.
The absurdity really does compound the deeper you go. Every layer of this onion is another bureaucratic compromise.
I think there's something almost comforting about that, in a strange way. The fact that the air marshal program is a messy, imperfect, slightly ridiculous human institution rather than a sleek, omniscient security apparatus — that's actually reassuring. It means the people running this program are grappling with the same constraints and trade-offs as everyone else. They're not omnipotent. They're just doing their best with a nearly impossible mission.
There's a practical step for listeners who want to go deeper on this, by the way. The FAMS FOIA documents on MuckRock are a rabbit hole worth falling into. You can see the redacted SOPs, the training manuals with entire paragraphs blacked out, the budget justifications with numbers that don't quite add up. It reveals the mundane reality of the program in a way that no news article ever could. You start to realize that behind the Hollywood myth is just a government agency with filing cabinets and interoffice memos and people arguing about per diem rates.
The most revealing document I've seen is a redacted after-action report from a training exercise where the marshal accidentally left his weapon in the lavatory. It's the most human thing imaginable. You're a federal agent, you're armed, you're supposed to be the last line of defense against terrorism, and you forgot your gun next to the sink. The report is heavily redacted, but you can still feel the embarrassment radiating off the page.
That's the real air marshal service. Just a guy who really hopes nobody noticed he left his service weapon in the bathroom.
Where does this leave us? Is the air marshal program a vital tool or an expensive security blanket? The honest answer is that it's probably both, and the ratio depends on how you weigh deterrence value against measurable outcomes. If you believe the deterrence is real and significant, eight hundred and five million dollars a year is a rounding error in the federal budget. If you're skeptical that a one point four percent coverage rate meaningfully deters anyone, it starts to look like a very expensive jobs program for people who are very good at sitting quietly in middle seats.
There's an open question about where the program goes from here. The twenty twenty-six TSA reauthorization bill includes a provision that would allow marshals to work remotely on certain flights — monitoring CCTV feeds from the ground rather than physically being on the aircraft. If that goes through, it could fundamentally change the nature of the program. The flying marshal, the guy in the suit with the gun, might become a thing of the past.
Which raises its own set of questions. If the marshal is on the ground watching a video feed, what exactly is their response capability? They can call the cockpit, they can coordinate with law enforcement at the destination airport, but they can't physically intervene. It's not an air marshal anymore. It's a remote security monitor. And that might actually be a more cost-effective model, but it's also a complete abandonment of the original mission.
With the rise of drone threats and cyber attacks on aviation systems, you could make a strong argument that the eight hundred million dollars would be better spent on counter-drone technology and cybersecurity for air traffic control systems. The threat landscape has shifted since two thousand one, and the FAMS program is still optimized for the last war.
That's the core tension. The program was built to prevent another nine eleven, and it may have done that. But the next attack on aviation probably won't look like nine eleven. It might not involve hijackers at all. It might be a drone swarm over a runway, or a ransomware attack on a regional air traffic control center, or something we haven't even thought of yet. And a guy with a gun in seat fourteen C can't do much about any of those.
The next time you're on a plane, look around the cabin. Statistically, there's about a ninety-eight point six percent chance that no one on board is a federal air marshal. But there's a one hundred percent chance that the flight attendants know what they're doing, that the cockpit door is reinforced, and that your fellow passengers have seen enough movies to know that you don't let the bad guy get to the front of the plane. That's the real security system. Everything else is theater.
Theater with an eight hundred million dollar budget and a code name that sounds like a GI Joe character.
Now: Hilbert's daily fun fact.
Hilbert: The oldest known cheese residue ever discovered was found on pottery shards from a Neolithic settlement in Hokkaido, Japan, dating back to approximately three thousand BCE. Analysis revealed it was a fermented fish-and-mammal-milk hybrid cheese, making it both the earliest cheese in East Asia and arguably the least appetizing culinary experiment in recorded prehistory.
Three thousand BCE fish cheese. That's going to sit with me.
I'm going to think about that for the rest of the day and I'm not happy about it.
This has been My Weird Prompts. Our producer is Hilbert Flumingtop. If you have a weird prompt you want us to dig into — and we mean weird, the kind of question that makes you look up from your boarding pass and wonder what's actually happening behind the scenes — send it to us at myweirdprompts.That's myweirdprompts dot com. We'll be back with another one soon.
