This week we wrote about Washington reopening the question of how much of the lethal decision it will hand to a machine. The hardware side of that question, it turns out, had already taken a concrete step days earlier. On July 16, at Eglin Air Force Base in Florida, DARPA and the U.S. Air Force flew an F-16 under the control of an artificial intelligence agent.

An AI flying a fighter jet is not, on its own, new. What makes this flight matter is which jet.

From a one-off testbed to the fleet

The Air Force has flown AI-piloted fighters before, most famously the X-62A VISTA — a heavily modified, one-of-a-kind aircraft that proved an AI agent could hold its own in a dogfight. VISTA was a laboratory with wings: bespoke, instrumented, not something you could ever field at scale.

This flight was different. It used a standard operational F-16, modified under a program called VENOM — the Viper Experimentation and Next-generation Operations Model, run jointly by the Air Force and DARPA under the longer-running Air Combat Evolution effort. The point of VENOM is precisely the thing VISTA couldn't demonstrate: that you can take ordinary fleet aircraft and give them autonomy, rather than building exotic testbeds one at a time. It's the step from "an AI can fly a plane" to "an AI can fly the planes we already own."

The design keeps a human firmly in the loop. Pilots can toggle between traditional human control and AI control, in DARPA's phrasing, "with the flip of a switch" — a person stays aboard, able to take back the aircraft instantly. This is human-on-the-loop testing, not an uncrewed weapon.

What the people running it said

"These groundbreaking flight tests of VENOM-modified F-16s advance the infrastructure needed to develop trusted, autonomous air combat capabilities," said Brig. Gen. James "Fangs" Valpiani, the DARPA program manager. The word doing the work there is infrastructure — VENOM is being built as the foundation for what comes next, not as a finished capability.

What comes next is named too. The VENOM fleet becomes the testbed for DARPA's Artificial Intelligence Reinforcements program, or AIR, which will fly multiple AI agents in live scenarios. "We now have the opportunity to create dominant autonomy for beyond-visual-range, multi-ship combat," said Lt. Col. Patrick "Dice" Highland, the incoming AIR program manager. Translated: the goal is human pilots commanding teams of autonomous aircraft — the "loyal wingman" concept, moving from slideware toward flight test.

Why it matters

Put this next to the policy story. Earlier this week the U.S. was visibly reopening the debate over how much lethal decision-making to delegate to software. That debate has always had a comfortable buffer built into it: the machines weren't really ready. Flights like this one narrow that buffer. The question of what a system may be allowed to do gets more urgent as the question of what it can do keeps getting answered.

None of this is autonomous weapons deployment. A modified F-16 flying a test sortie with a pilot ready to grab the stick is a long way from a machine choosing targets on its own. But it is the direction of travel, stated plainly by the people building it — and it's now flying on aircraft the Air Force already has in its inventory, not on a single science-project jet. The policy conversation and the flight-test conversation are converging, and the second one is moving fast.