AI War Revolution: Pentagon's $54 Billion Autonomous Warfare Blueprint

$54 billion will be spent by the U.S. Department of Defense on autonomous warfare technology, even as it remains undecided as to what these technologies are really authorised to do. Rather, it is an operational gap being widened at the very time that speed and scale are wiping out the decision hierarchies of war.

Illustration by The Geostrata

With the Defense Autonomous Warfare Group (DAWG) budget proposal for fiscal year 2026-27 and National Security Presidential Memorandum 11 (NSPM-11) coming into effect in early 2026, it is evident that the Pentagon is adhering to a doctrine that still needs to be fully formulated.

The reasons behind the approach are simple enough: autonomous technology holds the potential to multiply military strength, accelerate targeting, and distribute decision-making in challenges for environments where humans cannot work effectively. The issue of governance is just as simple, as there is no concrete operational definition of autonomy, authority for lethal use, and escalation management.

THE INVESTMENT AND THE DIRECTIVE

DAWG is not a futuristic initiative either. It is the very representation of changes in institutions that lead to taking autonomous decisions in the aspects of air, naval, and ground domains. The investment request includes the cost for developing the platforms, integrating the AI systems, coordinating swarm algorithms, and testing the environment. NSPM-11, though modest in its public statement, has basically authorized the autonomy program of the military.

What is important here is not the dollar figure itself, but rather how fast this is growing. The Pentagon is progressing from experiments with autonomous systems (like drones swarms and fighter aircraft teamwork) to actually operating systems that can target and attack in fully automated ways with minimum human intervention. The technological challenge is already solved. The political one remains.

As for the timeline: NSPM-11 published in spring 2026. During the summer, the services are working on targeting protocols. By 2027, the first autonomous system architectures were going from simulation to limited operation. The military is running out of time even faster than its policy-making process.

THE PROBLEM OF THE RULES OF ENGAGEMENT

In conventional Rules of Engagement (ROE), a human being is part of the decision chain. These rules outline who should authorize firing, when such a move is allowed, which targets it is acceptable to fire at, and how much collateral damage is acceptable. The commander receives the brief on the intended target, and the firing begins.

The autonomous system accomplishes all of that in software. An AI algorithm analyzes sensor inputs, recognizes the most likely target, computes the firing solution, and fires perhaps in just seconds. It’s the most likely part that’s the issue. Machine learning algorithms aren’t like military lawyers. They’re engineered for pattern recognition. They don’t make fine-grained distinctions between military positions and civilian buildings in an urban environment.

This is treated as a problem of scale by the documents within the Pentagon, increasing the ratio of human intervention, making it "human controlled" rather than automated. However, this does not address the fundamental problem. Human control is only nominal in times of high-intensity conflict. If the commander only has two seconds to decide whether to allow the system to shoot before it goes into a civilianly populated area, then human control is reversed. The recommendation of the system becomes the decision.

However, NSPM-11 cannot fix this problem either. It sets out some principles: machines should be used according to human intention, autonomous systems need human supervision, and the decisions to employ force should be left to humans. But these principles do not constitute any architecture. They do not decide whether or not an attack aircraft may attack an aerial target autonomously using radar information only. They do not determine whether or not an autonomous naval system may open fire at a fast attack craft without confirming that action to the operator.

WHY IT'S IMPORTANT IN THE CONTEXT OF GREAT POWERS

Neither China nor Russia is going to wait for perfect governance frameworks. Both of the states are developing autonomous warfare capacities, and both are banking on the fact that getting there first outweighs getting there right. “Intelligentised warfare” has been incorporated into the military doctrine of China, referring to integrated systems in which human decisions become submerged in the rhythm of operations.

The advantage of the United States in this area lies in doctrine rather than in technology. Military doctrine of the United States is still oriented on human decision-making and responsibility. Chinese and Russian advances are more amenable to the autonomy-first approach.

The problem for the Pentagon is that sustaining such an advantage means going slower and taking more risks than their rivals.

This is the genuine strategic gamble hidden in the $54 billion bet. What the Pentagon means here is that they can go faster and be more autonomous than they are happy about, and they are ready to do so since the other option would be to lag behind. It is a statement of faith regarding U.S. system design hiding huge uncertainty about governance.

INTERAGENCY SILENCE

What stands out about the interagency process is how much has not been done. There is little State Department participation in the governance of autonomous warfare. Congress is responsive rather than anticipatory. The international allies are engaged but not directing American architectural decisions. The global governance framework, to the extent it exists, does not seem to be restraining Pentagon planning.

It looks like the competition among militaries is outpacing the diplomatic and legal structures. By the time State and Defense have decided how they want to portray autonomous warfare against their opponents, the technology will be available. The framework will come after deployment rather than before. That is how arms control agreements are made to fail.

The Pentagon isn't careless. Military leaders know the dangers of escalating a situation. The DAWG project has a safety analysis component, along with simulation trials. However, there is an institutional issue here. The body responsible for creating weapons systems, namely the military, is not the same body able to govern their usage. The military can create the weapons system better, faster, and more intelligently. It cannot determine how it is used on its own.

The fact that a $54 billion investment will be made is a way of forcing a choice. This means that a decision must be made. One cannot remain in an ambiguous position where the possibility of using autonomous capability exists but the framework doesn't exist.

Choose your doctrine, limit the weapon system, and preserve human control, or move forward with automation and accept all consequences of that choice.

At the moment, the Pentagon wants to have its cake and eat it too. It’s talking about maintaining human oversight while developing technology that is optimized for machine decision-making. It’s allocating huge amounts of money without knowing what the conditions of use will be. This is where the risk exists.

---

BY HEMA

TEAM GEOSTRATA

info@thegeostrata.com