The MQ-20 Avenger has quietly crossed a threshold that once belonged only in speculative fiction, using onboard artificial intelligence to line up and execute a simulated air-to-air kill without a human pilot in the loop. The latest test capped a multi-year campaign to prove that an unmanned combat jet can not only fly itself, but also hunt, track, and engage enemy aircraft while respecting strict safety and command constraints. It is a glimpse of how future air wars could unfold, with software making split-second decisions that used to rest solely with fighter pilots.
What makes this milestone different is not just that the MQ-20 hit a virtual target, but that it did so as part of a broader push to turn the Avenger into a fully networked, combat-ready teammate for crewed fighters. The simulated kill sits alongside live intercepts, beyond-line-of-sight trials, and complex teaming experiments that are steadily turning the jet-powered drone into a central player in U.S. concepts for Collaborative Combat Aircraft.
The test that proved an AI jet can “kill” on its own
The most recent demonstration of the MQ-20’s autonomy paired a live aircraft with a virtual opponent, allowing the drone’s onboard systems to manage a full air-to-air engagement from detection to simulated missile shot. According to program descriptions, the engagement used a detailed simulation to determine that, had a real weapon been fired, the intercept geometry and timing would have resulted in destruction of the target, a level of fidelity that moves the event beyond a simple lab exercise and into the realm of operationally relevant testing. The same trial showcased how the Avenger’s autonomy stack can work with external partners such as Lockheed Martin and L3Harris Technologies, integrating their systems into the engagement chain and validating the simulation as a credible stand-in for live weapons.
That simulated kill did not emerge in isolation. It builds on a broader portfolio of work by General Atomics Aeronautical Systems, Inc., which has been steadily expanding the MQ-20’s role from a technology demonstrator into a core element of future force design. The company positions the Avenger alongside its Predator and Reaper families as part of a continuum of unmanned systems, but with a jet-powered airframe and stealthier profile tailored to contested airspace, and it has used its corporate platform to highlight how the MQ-20 fits into emerging U.S. Air Force and Navy concepts for unmanned combat aviation through its main GA-ASI portal.
From first simulated kill to live intercepts
The path to this latest engagement runs through a series of earlier trials that gradually handed more responsibility to the aircraft’s artificial intelligence. In an early February campaign, Shield differentiated between the first flight test of a Hivemind-enabled MQ-20 Avenger UCAV during the Orange Flag exercise and later work, noting how the February test focused on validating basic behaviors while human controllers monitored the unmanned combat aerial vehicle’s reactions to dynamic air-to-air scenarios. That initial outing of the Hivemind software on the Avenger UCAV at Oran paved the way for a later event in which the AI achieved its first simulated air-to-air kill, a milestone that marked the MQ-20’s transition from experimental autonomy to a credible combat agent, as detailed in reporting on the February test.
By mid-2025, the focus had shifted from purely virtual engagements to mixed live and synthetic environments. In one high-profile campaign, the company described a “Newest Groundbreaking GA-ASI Autonomous Jet Demo Includes Successful Simulated Shoot-Down,” in which the Avenger’s autonomy stack coordinated with live and virtual aircraft to prosecute a target and execute a simulated shot against live targets. That event, highlighted in the Autonomous Jet Demo description, showed that the MQ-20 could function as part of a complex kill web, sharing data and acting on it fast enough to matter in a real fight.
Shield AI’s Hivemind and the brains behind the Avenger
At the core of the MQ-20’s recent achievements is Hivemind, the autonomy stack developed by Shield, which is designed to let unmanned aircraft make tactical decisions at machine speed while still operating within human-defined rules. Shield has described how Hivemind enables the Avenger to plan routes, react to threats, and manage engagements without continuous remote piloting, a capability that was first proven in the February Hivemind-enabled flights and then refined through subsequent simulated kills. The company’s own overview of its technology emphasizes that Hivemind is intended to deliver combat-ready autonomy across multiple platforms, a vision that underpins the Shield roadmap for future air combat.
Shield AI has also framed its work on the MQ-20 as part of a broader push to field operationally relevant autonomy, not just experimental algorithms. In a later update, it highlighted how Shield AI Advances Combat-Ready Autonomy with Second MQ-20 Avenger Flight, describing a second Avenger Flight from SAN DIEGO that focused on refining Hivemind’s performance in realistic mission profiles and demonstrating that the software could handle more complex scenarios without sacrificing safety. That second MQ-20 Avenger Flight, detailed in the Shield AI Advances release, underscored the company’s intent to move quickly from lab demos to fieldable capability.
Company-funded demos, Navy teaming, and adversary roles
General Atomics Aeronautical Systems, Inc has not waited for formal programs of record to mature before pushing the MQ-20’s envelope. In a Company Funded Demo described as “GA-ASI Makes Another Autonomous Aerial Intercept in Company-Funded Demo with MQ-20 Avenger,” the firm detailed how the Avenger executed another autonomous aerial intercept while adhering to operator-assigned Keep-Out Zones and Keep-In Zones, demonstrating that the aircraft could aggressively pursue a target while still respecting geographic and safety constraints. The MQ-20 showcased adherence to those KOZ and KIZ boundaries throughout the mission, a key requirement for any autonomous system that will share airspace with crewed aircraft, as highlighted in the Keep-Out Zones description and reinforced in the broader Makes Another Autonomous account.
The MQ-20 has also been pulled into formal experimentation with the U.S. Navy and Air Force. In one demonstration, GA-ASI and the US Navy flew the Avenger using an MD-5 Ground Control Station to perform commanded autonomy maneuvers, an event explicitly tied to advancing technology for future Collaborative Combat Aircraft and the U.S. Air Force’s CCA program. That trial, described in the Collaborative Combat Aircraft overview, positioned the MQ-20 as a pathfinder for how uncrewed jets might be directed by ground operators while still exercising significant onboard autonomy. In parallel, the Pentagon has tapped the Avenger for Project Red 5, which revives a US Air Force adversary air training program that had been canceled in 2023 and had its funds diverted to the Collaborative Combat Aircraft program, a move that gives the MQ-20 a role as a high-end sparring partner for frontline fighters, as outlined in the Project Red reporting.
From Orange Flag to beyond-line-of-sight and live intercepts
The MQ-20’s maturation has been shaped by a series of increasingly demanding exercises. During the Orange Flag exercise, General Atomics Aeronautical Systems, Inc equipped the Avenger with U.S. government-furnished autonomy software and Shield’s Hivemind AI to explore how the jet could handle air-to-air engagements with minimal human input, a testbed that allowed engineers to observe how the system behaved in a complex, multi-asset environment. That Orange Flag work, described in detail in the Orange Flag account, set the stage for later trials in which the MQ-20 would not just participate in exercises but take the lead in executing engagements.
More from Morning Overview