The Future Of Warfare- Collaborative Combat Aircraft And AI Autonomy.mp4
Nov 2, 2025
Hi, I'm Tracy LaTourette, Colorado's First Lady fighter pilot.
Call sign, Jackie O.
We're here at Wings Over the Rockies Air and Space Museum in Denver, Colorado.
We're getting an inside look at an entirely new type of wingman
called a Collaborative Combat Aircraft, or CCA.
We've gone from the lab to the simulator to this aircraft that we can fly right alongside.
It was this moment in the history of aviation that has never happened before.
As a test pilot, I'm naturally skeptical of all new technology.
It's a very messy, ugly, complicated, confusing endeavor.
Let's go test.
On test day, there's nowhere busier than mission control.
What we're doing is going to change air power.
Launch, launch, launch.
The idea of robots that could kill people, being able to do that without supervision, deeply concerning.
Human beings are always going to have to have a role in overseeing it.
I think everything's at stake.
It's time to go behind the wings.
Above the white, sandy beaches of Eglin Air Force Base in Florida, a jet screams by.
But there's no pilot.
In fact, there's not even a cockpit.
This is the XQ-58.
It's an uncrewed tactical aircraft using artificial intelligence and autonomy
that could revolutionize our air battlefield.
Artificial intelligence is rapidly changing the face of our air battlefield.
Take us back a bit.
How did we get to where we are today?
Artificial intelligence has actually been around for decades.
But more recently, we've had the confluence of big data, compute power, and algorithms coming together.
It is really a revolutionary moment that I think has forever changed so many aspects of our life, including aviation.
We used to have deterministic software code.
Deterministic software code would be those if-then statements.
If this happens, then this is going to happen.
And you would hard-code everything in.
It was great in the sense that it was repeatable.
So it would always have the same output.
But now, you're at a point where these machines don't have to be hard-coded anymore.
Because at times, the challenges that we face, there are too many ifs to give us the thens with those if-then statements.
So now we need artificial intelligence to help us account for the unknowns that the data possibly has and then to adjust itself.
Now it's not as repeatable, but it's more responsive and agile.
And it's the next evolutionary step in software development.
We're talking about military applications that can be potentially lethal.
What's really at stake here?
I think everything's at stake.
We have got to approach this technology with that ethical mindset.
A human has got to be at the center of any lethal decision-making.
Just like our Predator uncrewed aircraft have an operator on the ground making the decision when to drop or fire something that could have lethal effect.
The same thing is going to be true for any type of AI-enabled systems.
We cannot afford to get those things wrong.
Frag, you're no stranger to the combat zone.
Part of your job now is looking at the Air Force and what we need in the future.
Where did this idea of collaborative combat aircraft even come from?
One of the things that we do is we do things that we call war games.
We set up scenarios that look at particular environments, particular objectives that a combatant commander needs to achieve or problem sets that they need to do.
We set these scenarios up and then we set up both a friendly side of the force and then an opposing side of the force.
We bring experts from all across Department of Defense and some parts of the interagency in to then conduct what we call this war game.
Over the last decade as we've conducted these different war games, we start to identify some pretty clear capability gaps.
Things that we don't have the equipment and so once we identify those gaps, we realize that we need to start getting after what can fill that gap.
And in one of these war games, years and years ago, we identified if we had more mass and we had it in more affordable ways
and we started moving into autonomous behaviors to alleviate human workload and risk, then it provided us a tremendous amount of capability.
I'm excited about the XQ-58 Valkyrie. I think it does represent a change in what we can do with aircraft.
Whether that technology can go the next step to get you to not just a loyal wingman but the actual fighter pilot itself,
I don't know if it can make that jump. I think there's quite a gulf there between the two.
I would be on the side of going slower rather than faster.
We have to start worrying if we don't use those capabilities, will our adversaries use those capabilities?
Do we have to put them in just because we know that they're going to be moving faster than human decision speed?
And it's that back and forth, that potential for an AI arms race, that I think puts us at a really interesting tipping point for AI right now.
I don't think we need to rush things out because, remember, the arms race works both ways.
If the adversary believes we are rushing things out too early, they will be incentivized to rush things out even earlier, and that's bad for everybody.
I think maybe by going more deliberately slower and making sure that the AI is responsible and ethical on our side,
that may lead us to a better future than if we do it the other way around.
The United States has made rapid advances in this kind of technology,
but what happens if an adversary starts deploying this kind of technology sooner? What do we do?
We're going to be at a disadvantage, and I think we'll have to rush at that time to try to get it into production and building quicker.
I think it's important that we limit our reach for our first increment in particular,
that we get the set of capabilities that are most important to us, that give us the greatest increment of advantage over an adversary.
The way the program is structured today for the uncrewed collaborative combat aircraft is in two increments right now.
The first increment gets substantial numbers fielded within the next few years,
but then right behind that is the next increment that will have more advanced technologies, additional functionality,
and another level of capability compared to the first one.
We are building up on real integration of all those capacities into the military,
so the moment is immediate in terms of being able to engage on it from a civilian outsider perspective.
The idea of robots that could kill people, being able to do that without supervision,
or the human control that has always gone into weapons and war, deeply concerning.
The Pentagon has been asked directly, do you intend to empower these drones to deploy weaponry?
Are they autonomous weapons? And the answers have really been hedging.
Are we always going to have a human involved in that kill chain?
Are we ever going to get to a point where the machine is actually making that lethal decision?
There are questions as to exactly when in the chain and how much of that decision-making process is delegated as subject to the machine.
At the end of the day, we will always have humans involved in the application of force.
Humans who put that machine on the battlefield, who supervised it, who designed it, are accountable for its performance,
and we've got to maintain that. I think that's central to the rule of law, basically, and to ensuring compliance.
And I think, frankly, as we get into this, we'll see machines that are better than humans
at both doing the operational tasks they've been asked to do and at following the rules of engagement.
It starts early with the technologists, those good ideas inside the places like the Air Force Research Lab
or Defense Advanced Research Projects Agency, DARPA.
You'd be able to evolve some of those ideas, ultimately come into something into a physical nature, like the XQ-58.
Originally, that vehicle was built and flown with the idea of how can we manufacture a low-cost vehicle
to meet the affordable mass component. We took that good idea, we iterated it over time,
we tested it with the XQ-58 and the X-62 out at Edwards Air Force Base,
and then ultimately it becomes a fully funded program and says, no, this is a technology we need
as part of our baseline Air Force to be able to go ahead and meet the joint requirements.
We build and deliver over 160 aircraft a year.
Most of those are target aircraft, an aircraft that has the performance of a Valkyrie,
but at a tiny fraction of a manned aircraft cost.
And when you talk about affordability, what are we talking?
The typical manned aircraft fighter-type systems, which are, you know, pick the number,
$50 to over $100 million as compared to a missionized system, much less than $10 million.
After we had these target aircraft, we started looking at the U.S. military need.
We have these aircraft systems that are used as targets today,
but all of their airborne characteristics match the characteristics of advanced fighters.
We said, you know what, we'll change the mission systems, so think payloads,
think about avionics that go on the airplane.
We'll change some of those out on a target aircraft.
A target aircraft we shoot at, a tactical aircraft system like the Valkyrie shoots back.
We're standing here in front of the XQ-58.
The Valkyrie weighs 2,500 pounds.
It's 30 feet long and 27 feet wide.
It has an 1,800-pound payload when you count the 600 pounds it can carry internally
and 600 pounds under each wing.
Cinco, introduce that.
This aircraft right here, as you can see, has no cockpit, so it's an uncrewed aircraft.
It's what we consider high performance, so it is able to go transonic,
so not quite over the speed of sound but really close to the speed of sound.
It's actually larger than most people imagine,
and it's around the size of what we consider around a T-38,
which is an aircraft that we use to train fighter pilots.
So it's very similar in that sense.
You see the engine inlet is in the top right there,
and the aircraft itself is actually not as important as people imagine.
What is important that's going on in here is what's on the inside.
The autonomy functions and the AI-enabled autonomy functions
that we're testing out with our software and with different aspects of that software
interacting with the different control surfaces, that's what this aircraft is all about.
So it's uncrewed, it's super high tech, but it's also pretty straightforward in a lot of ways.
Can we take a look?
Yeah, absolutely.
You know, so it really is quite simple.
It has this complex aspect of it, of the insides being like this AI agent,
futuristic, high-performance, uncrewed aircraft,
but then the outside is not a high-G force air vehicle.
It's not stealthy at all.
This is an experimental testbed, and we're testing the innards.
How does the autonomy work with other systems?
How does the communication work with the ground station, the flight controls?
So that aspect of it, allow us to put it in scenarios that are representative to fighter aircraft,
which then allows us to develop that type of capability.
So there's no landing gear.
We don't need a runway.
It just launches.
What kind of engine does it take to make that happen?
I'll show you right now.
Compared to an F-35 or an F-16, it's kind of cute.
Like, it's little.
And so this actually doesn't launch it.
This engine is running, of course, when we launch it, but this is the launch rails that it's on.
So this thing actually rotates up, and there's rockets that are on the side of it,
and those kick off, and this thing rockets off the rail.
Those rockets fall basically immediately to the beach here,
and the aircraft is at that point able to fly.
It has lift, it has thrust, and it's able to go do its mission.
As we are developing what the actual airframe is going to look like,
this is just meant as an experimentation bed to help us get to that end state.
It has parachutes that pop out the top, and it's able to land on our range up here.
And then we're able to reuse the aircraft, possibly days later if needed.
So it provides us with a unique and different approach to how we take off and land,
but that's not necessarily how these are all going to be moving forward.
The XQ-58 is really a pretty simple aircraft but has an extremely complex mission.
I'm so excited to see where it takes us in the near future.
We've gone from the lab to the simulator to this aircraft that we can fly right alongside.
Tell me about the simulation and testing that has gotten us to a place
where you can trust flying alongside an uncrewed aircraft like this.
As a test pilot, I'm naturally skeptical of all new technology
because I want to see it proven to work the exact way that we expect it to.
We went through thousands of hours of simulation
for the software that was going to be running on this air vehicle.
It is super important that we trust exactly how this will operate
when it goes time for me to fly alongside it.
It'd be great to just show up, hop in your F-16, and go fly alongside the XQ-58,
but we both know there is a lot more that goes into the mission.
As a test pilot, what does a typical sortie look like working with an AI-enabled platform such as this?
So we start months in advance figuring out what exactly objectives we want to test.
Once we've defined the path forward, finally that all comes together the day of flight.
We show up super early and then we prepare the aircraft for launch.
We'll go through a pre-flight briefing to make sure that we are aware of all the factors
that could affect the flight for that day, and then we go fly the flight.
Each mission starts with a flight brief where we go over the plan for the day.
All right, good morning everyone. My name is Slap.
I'm going to be leading DEVIL flight today for the safe, effective, efficient, and flight test of the XQ-58.
There is no alternate mission for today.
For our takeoff, expect that our fourth-generation chase aircraft are going to take off from Eglin Air Force Base.
Torch, you briefed Northflow Central Control Facility.
Folks, you guys owned the go-no-go decision for XQ-58 launch today.
I expect to hear from you a clear countdown on the radio,
and we'll position ourselves to immediately execute an airborne pickup and rejoin on the XQ-58 as it launches.
For a test point progression today, we're going to start with basic air vehicle maneuverability,
and then we're going to go on from there into some of our AI testing, okay?
Starting with limits and then working on to challenge problems after that.
For Set 3 and 4, those will be different AI agents.
We'll be executing problems Victor and Whiskey from Set 3, and Set 4, number Alpha and Echo.
Torch, you and I are now going to go get our gear on, and we're going to head out to our airplanes.
Vicky, you and Data are going to take the mission materials to the CCF, and then we'll see you on the radio.
Remember, our objectives today are safe, effective, and efficient flight test of the XQ-58.
Let's go test.
On test day, there's nowhere busier than mission control.
Eglin's CCF, which is the Central Control Facility, conducts over 1,200 test missions every year.
Here at the CCF, we have a team of test conductors, test specialists, safety specialists, and range and aircraft controllers
that work together to make sure we have a safe, secure, effective, and efficient mission.
On a launch day, this place must be packed. What's the energy like?
Every seat is filled in this room, and we're in constant communication with the people at the launch pad.
It's excited, but also subdued, and we make sure we go through all our pre-flight checks,
make sure the air vehicle is ready to go, and then when all of those are finished, we've finished our checklist,
then we begin the countdown to the launch.
I love my F-16, but a rocket-assisted takeoff? That's pretty awesome.
This is the part I've been waiting for. Let's see this thing fly.
This is Project, continuing pre-launch checks. Launch time will be 1420 Zulu. Chase ready?
Chase ready.
Green range?
Green range.
Recovery offset?
Offset.
Winds?
Wind limits.
All right. Proceed with engine start.
Copy engine start.
10, 9, 8, 7, 6, 5, 4, 3, 2, 1, launch, launch, launch.
It is quite a roar when the rockets actually ignite and take off, but it's a very brief roar,
and in those few seconds, you know that this aircraft is going somewhere.
Radar kit separation.
Everyone has a role to play. Everything is all about safety. Ranger has control.
It has to be pretty cool flying alongside this aircraft, but you're not up there just enjoying the flight.
What are you doing exactly?
I'm communicating with the ground controllers and my test conductor
to make sure that we're running the test points efficiently,
but there are those moments where I look outside and it's just nice to soak it all in.
Flying alongside this aircraft for the first time was eerie.
There is no pilot in the cockpit controlling it.
It maneuvers a little bit differently under AI control than a pilot would.
It's not limited by those human factors that limit a crewed aircraft,
so roll rates and maneuvers happen a little bit more rapidly sometimes and in unexpected ways.
I am constantly on alert for that
and constantly making sure that the airplane is still performing as expected.
It's much more than just day of.
It's a lot of study, it's a lot of hard work,
and it's a lot of teamwork from the folks here at Eglin to make a successful test flight.
For this mission, we have two aircraft flying chase,
so they are making sure that the air vehicle is flying in a way that is expected.
We also have engineers here in the control room that are looking at the telemetry data
and we can tell what the agent is doing,
and if it trips off, we can immediately assume control with human remote control operators here at the CCF.
It is a really long day, but from the very beginning with the pre-launch checks to launch all the way to recovery,
our job's not done until we have the vehicle back in the hangar at the end of the day.
XQ in the chutes.
Two airbags visual.
And touchdown.
Great job, everybody.
Really what we're doing is going to change air power.
With each test of the XQ-58,
we get a more vivid look of what the air battlefield could look like with operational AI.
I understand you were there the first time an AI agent flew a high-performance uncrewed aircraft.
What was that like?
So I had the honor of flying on the wing of this aircraft the first time an AI agent piloted a high-performance uncrewed aircraft.
It was this moment in the history of aviation that has never happened before.
So there I was in my Strike Eagle, a 1980s bird, flying next to this uncrewed aircraft.
No pilot, no cockpit.
It was fantastic to see the culmination of all the work that was put in from the research at MIT and other institutions
all the way into the Air Force Research Lab, experimenting with the test beds at test pilot school.
And then there I was.
And it was amazing because it was forever going to change aviation.
You don't see our job going away anytime soon, do you?
I think we need to reflect on the past.
As aviation has progressed, the role of the pilot is different.
When I jumped in the F-35 and was commanding that amazing vehicle,
I found myself actually letting the aircraft fly much more than I was used to.
I would turn on auto throttles, so the engines basically would keep an airspeed that I set.
I would turn on autopilot, so it would do a lot of the flying functions
where then I could focus on my touchscreen in front of me in conducting the mission,
kind of being the quarterback of the airspace.
And this is the next iteration, I think, of what the future of aircrew looked like.
We have a huge responsibility in front of us.
Artificial intelligence-enabled autonomy is changing everything.
And so how we approach it, how we test it, how we develop it, how we implement it
is of ultimate importance.
The XQ-58 Valkyrie can be used for a supporting mission
where it's flying alongside human pilots.
The technology in it could be used in terms of surveillance and reconnaissance.
It's getting to the point where it can be a real supporter of other combat aircraft
and hopefully help our fighter pilots survive in an environment they might not have been able to otherwise
when they're all alone.
With the right controls, it could be used to launch weapons,
but those are much more constrained instances with that.
Usually we would want the human to check that for now.
We're not prepared to give up that level of autonomy, nor should we yet.
The scope is going to be anywhere and everywhere.
Multi-domain is what the real target is, is to be able to integrate this
in not only one element of the U.S. military, but nearly every element.
So I think there are people who are trying to use it in aviation, trying to use it on sea craft.
But I think in mass swaths, it is a force-wide operation.
There is a very inherent moralism that goes into decisions about life and death and war fighting.
That moralism is sometimes stripped out when people who are making these decisions
believe that, one, a system is making that decision better than they possibly could.
Two, the system is trustworthy.
And three, they don't have blood on their hands if the system created that decision.
We can talk all day about the particular systems,
but really the thing that we're concerned about as an outside advocacy group
is how does this block off people's morality in creating something like this and making it.
The main way that we're fighting war means that we may be turning off
some of those critical thinking capacities.
We may be stopping places in the chain of command
where you would otherwise have common sense judgment come into play or a decision changed.
And removing all of that is terrifying.
There are risks involved with this kind of technology,
but what are the risks inherent with not pursuing the development of AI and autonomy at all?
To AI or to not AI, that's the question.
The amount of data that's coming in for a human commander to make decisions on
will be at speeds at which they'll need to look at a list of options
that are being analyzed by AI machine learning algorithms to present options.
To not do it means that you will be behind in decision making.
The enemy will be making decisions faster than you on better data
and have different options in a very complex environment.
The other part is if we don't pursue this and potential adversaries do,
they're going to have an inherent edge.
And we already know that most of our peer adversaries around the world
are going to pursue this technology.
So we have to embrace it. We have to move forward.
And it's going to enable us to do things better, more efficiently,
and actually at the end of the day save lives, not the other way around.
So you have aircraft like the XQ-58 that are testing regularly now.
What kind of things are we learning and where are we going next?
We break the system down in a few different areas.
The first one being the autonomy piece. We have to get that part right.
The second piece being the actual air vehicle, which is the next piece.
And then finally making sure we can understand operationally
how we're going to employ this system.
Those are the three areas we're ideally focused on.
And so the XQ-67 is yet another variant by one of our industry partners
that is allowing us to test another vehicle out inside our ecosystem.
The XQ-67A's major difference is the addition of landing gear.
We're moving into a new era right now.
And as you'll see with the XQ-67A,
we started off with a low-cost, shreddable platform.
And from there it's gone into what we call off-board sensing station,
which is the first of a series of airplanes.
We just flew our first off-board sensing station, XQ-67A.
They're helping enable the way towards the CCA set of programs that we have going forward.
We're not completely there yet. We've got a lot to learn.
In fact, we will never stop learning on this.
I predict that you're going to see collaborative combat aircraft sooner than later.
There will be a lot of them out there.
And things are moving quick right now.
And they're moving quick because the Air Force is demanding that we move quick.
Where do you see it going soon?
And where do you see it going over the next 10 to 20 years?
We need to talk about the specific applications.
And whether the technology is matured enough to take on those applications.
We're moving forward very quickly on collaborative combat aircraft
because we think that that set of applications is ready and we can go ahead and move forward.
There are going to be others.
I don't think warfare will ever be entirely automated.
It's a very messy, ugly, complicated, confusing endeavor.
And human beings are always going to have to have a role in overseeing it.
We can only handle so much in our minds and we can only process it so fast.
And the machines are going to do better.
But they've got to be accurate.
And they also have got to be consistent with our values and our ethics and our legal obligations.
As testing progresses, artificial intelligence-enabled platforms like collaborative combat aircraft
are closer than ever to being operational.
We'll see you next time on Behind the Wings.

