The U.S. Air Force has been flying its fighter drones for over a year. It has tested them, taxied them, launched them from austere strips and flown one alongside an F-35. What it has never done is put them into a proper war.

In December, it will. Collaborative Combat Aircraft are going to Emerald Flag, and for the first time they will be part of a large-force employment exercise rather than a test card.

The officer running that effort described it, with some understatement, as dipping a toe.

Why an Exercise Is a Bigger Step Than a Test Flight

A test flight answers whether the aircraft works. An exercise answers whether anybody can use it.

Emerald Flag has been run by the 96th Test Wing at Eglin since 2020, and it has always leaned heavily on networking and command and control rather than on shooting things. That makes it the obvious place to find out whether a drone can be told what to do by a human in another aircraft, in a crowded sky, while everyone is busy.

Bartlett was more specific about what will be attempted: the first live end-to-end demonstration of a CCA kill chain, and the first time the drones are folded into a combined fourth and fifth-generation offensive counter air push. His own assessment was that it will be small.

The Point Is Trust, Not Technology

The interesting thing about Bartlett’s framing is what he thinks the exercise is actually for. Not proving the datalink works. Proving that a fighter pilot will believe it does.

This is the recurring theme whenever American officers talk about autonomous aircraft. The engineering problem is largely solved. The problem of a human being willing to hand a weapon-carrying machine a task and then look away is not, and it does not get solved in a simulator.

Which is not to say the simulator is useless. Bartlett described a feedback loop where the operators sit with the programmers and changes land immediately, or on the next software build. That is how the autonomy gets good. The exercise is how it gets believed.

The Surprise: They Are Easy to Look After

Buried in the same panel was a number that matters more than most of the programme rhetoric.

One maintainer per aircraft. A modern fighter needs a great deal more than that. The unit also found that essentially any Air Force specialty can be trained to generate sorties and load weapons on these aircraft, because the machines are simpler and their job is narrower.

That is the whole economic argument for CCAs in one sentence. If the aircraft is cheap and one person can turn it, you can put it somewhere with a runway and very little else. The Experimental Operations Unit has already run two rounds of Agile Combat Employment testing this year, at Edwards in April and Creech in July, working on exactly that.

General Atomics footage of YFQ-42A flight testing conducted with the U.S. Air Force.

What Is Not in the Exercise

One clarification worth making. Boeing’s MQ-28 Ghost Bat already flew in a large-force exercise this year, Valiant Shield 26, and the Experimental Operations Unit was involved. But the MQ-28 was built for the Royal Australian Air Force and is not formally part of the U.S. Air Force’s CCA programme.

Emerald Flag in December is the first time American CCAs, the FQ-42 and FQ-44 the Air Force actually intends to buy, go into that environment. The service confirmed in June that its initial operational fleet will be a mix of the two, and it now wants at least 500 of them by 2032.

Five hundred aircraft is a serious air force. It is worth remembering that the first time any of them will fly in a proper exercise is three months from now, and that the officer running it expects it to be small.

Sources: U.S. Air Force; Air & Space Forces Association 2026 Air, Space & Cyber Conference; TWZ; General Atomics