Northrop Tests Mid-Flight AI Swaps on Talon IQ Aircraft

A modular autonomy test for military aviation

Northrop Grumman says it has demonstrated a new level of flexibility in airborne autonomy by swapping between different AI systems while an aircraft was already in flight. The company conducted the work on its Talon IQ testbed, a program previously known as Beacon, using a modified Scaled Composites Vanguard Model 437 configured for optionally autonomous flight.

According to Northrop and its partners, the recent tests showed that software from multiple companies could take turns controlling the same aircraft without interrupting safe operation. Northrop named Shield AI, Accelint and Applied Intuition as participants in the demonstrations. In one March test, Northrop said its Prism autonomy system handed full control to Shield AI's Hivemind after takeoff. In a later April flight, Prism remained in charge of the larger mission while handing control to Applied Intuition and Accelint for specific skills.

Why the handoff matters

The central claim from the companies is not simply that AI flew the aircraft, but that different AI packages could be swapped in during a mission. Northrop described the enabling technology as a layered, modular open architecture that separates baseline flight-control functions from higher-level mission autonomy. That distinction is important. The flight-control layer handles the split-second work that keeps the aircraft stable, while mission software determines what the aircraft should do, where it should go, how fast it should fly and how it should behave in a given scenario.

That architecture could make it easier to test, compare and deploy autonomy software from multiple vendors on a common platform. Rather than rewriting an entire control stack for every airframe, developers could focus on mission behaviors and specialized capabilities. Northrop presented that approach as a way to plug in either narrow tools for a particular task or broader systems capable of running larger parts of a mission.

What the tests included

Northrop said the latest flight was its eighth Talon IQ test. In the demonstration with Applied Intuition and Accelint, Prism reportedly delegated functions such as a Combat Air Patrol behavior while retaining oversight of the broader mission. In the earlier test with Shield AI, Hivemind took control after the aircraft was airborne and ran maneuvers including Combat Air Patrol and simulated target engagement maneuvers before returning control.

Those details suggest the company is trying to prove two related points at once: that autonomy can be portable across systems, and that command can be shared between multiple autonomy packages depending on what a mission requires. For military planners, the appeal is obvious. A common aircraft could potentially host different mission brains for surveillance, patrol, or engagement workflows without needing a new platform for each concept.

What comes next

The tests do not by themselves prove that autonomous combat aviation is ready for operational deployment, but they do show where the industry is heading. The emphasis is moving beyond one company building one closed AI stack for one aircraft. Instead, contractors are pushing an ecosystem model in which autonomy becomes more like software infrastructure, with room for multiple vendors and rapid iteration.

If that model holds up in more demanding testing, it could reshape how the Pentagon buys and upgrades autonomous systems. A modular aircraft architecture would let operators swap behaviors faster, compare suppliers more directly and avoid locking missions to a single software provider. For now, Talon IQ is still a testbed. But Northrop's latest flights point to a future in which the most valuable military airframe may be the one that can host many different decision-making systems, not just one.

This article is based on reporting by Breaking Defense. Read the original article.