The Drone Isn't the Weapon — The Software Commanding It Is | CSIS Analysis

The Drone Isn't the Weapon — The Software Commanding It Is

CSIS researchers Kateryna Bondar and Matt Mande, in their new brief Defining Autonomy: Why Software, Not Drones, Will Determine the Next War, argue that U.S. Department of Defense Directive 3000.09 (DoDD 3000.09) contains a fundamental blind spot in how it defines lethal autonomous weapons systems (LAWS). The directive classifies systems based on the behavior of the munition itself — focusing on the effects end of the weapon — while ignoring the decision-making layer above it.

What Does That Actually Mean?

Today's AI orchestration software can fuse data from multiple sensor streams, allocate targets, and coordinate fire sequencing across hundreds of platforms simultaneously. In functional terms, this software is already performing the "select and engage" decision. In other words, true autonomy resides in the software commanding the operation — not in the munition that flies out.

The two authors emphasize that excluding this "decision layer" from regulatory definitions represents a critical policy gap and a failure of accountability mechanisms.

Ukraine Provides a Clear Real-World Case Study

Ukraine's Delta battlefield management system was not built top-down according to procurement theory — it evolved bottom-up from frontline operational requirements. The system integrates drone imagery, intelligence, surveillance and reconnaissance (ISR) data, electronic warfare inputs, and allied intelligence feeds into a single common operational picture.

When software-directed drones now account for 80% of strike missions, that software is the war system. Taken further: whoever controls that decision layer can win engagements faster.

The U.S. Is Moving in the Opposite Direction

The U.S. military services are currently building competing orchestration software architectures in parallel, with each branch solving problems only for its own platforms — deepening fragmentation. For example, the Maven Smart System was designated in March 2026 as the backbone of Combined Joint All-Domain Command and Control (CJADC2), yet individual services continue developing their own systems concurrently.

Bondar and Mande consider this deeply problematic. They put forward five corrective principles:

1. Government ownership of the orchestration layer
2. Integration into training systems from day one
3. Genuine vendor neutrality
4. AI assessment mechanisms built into the system
5. Immediate experimental validation

The authors note that the "Drone Dominance" initiative and the Defense Autonomous Warfare Group (DAWG) are currently the Pentagon's best institutional vehicles for implementing these objectives before the next conflict erupts.

Conclusion

The United States is placing its bets on acquiring weapons at the effects end, while adversaries are already gaining the lead in controlling the software that commands those weapons.

In the twentieth century, the nation that built nuclear weapons first held a generational strategic advantage. In the twenty-first century, the nation that first builds the autonomous warfare orchestration layer will hold a comparable advantage — measured not in explosive yield, but in the speed, scale, and tempo of action.

This article was originally published in the Small Wars Journal at Arizona State University. The CSIS brief Defining Autonomy: Why Software, Not Drones, Will Determine the Next War was co-authored by Kateryna Bondar and Matt Mande.

原文來源： 查看原文