On September 30 at Quantico, Secretary of War Pete Hegseth announced something that deserves far more attention than another Pentagon reorganization.

Among six initiatives unveiled in his “State of the Force” address, Hegseth announced plans for a four-star Autonomous Warfare Command, or AUTOWARCOM, intended to build and scale autonomous and robotic capabilities across the Joint Force by October 2027. Project Agincourt, whose name evokes Henry V’s 1415 victory over a far larger French army, will lay the groundwork with new military specialties and faster acquisition pathways. Project Meridian will examine what technologies and capabilities future battlefields will demand.

“Lasting change must be anchored in military command structures,” Hegseth said. That sentence explains why Quantico matters. Experiments come and go. Commands endure.

Nine months earlier, at SpaceX’s Starbase, Hegseth declared that America would become an “AI-first warfighting force across all domains.” America would not win by bolting artificial intelligence onto old tactics, he said: “We will win by discovering entirely new ways of fighting.”

Quantico moves that vision from strategy to institution.

There are compelling reasons for doing so. The Pentagon’s latest China report finds that Beijing continues to develop military AI, and the People’s Liberation Army (PLA) is integrating AI into unmanned systems, cyber operations, and other military applications. America cannot assume potential adversaries will wait while we debate the implications. But history also warns against confusing technological superiority with military superiority. American forces have prevailed against technically superior weapons, such as Germany’s Tiger tank, because training, doctrine, logistics, mass, and human judgment also decide battles.

The harder question is whether we are building dependence on autonomous warfare faster than we are building the institutions capable of governing it.

Hegseth’s January AI Acceleration Strategy reaches across warfighting, intelligence, and military administration. Among its seven priority projects is Agent Network, intended to develop AI-enabled battle management and decision support extending from campaign planning to what the Pentagon calls “kill chain execution.” Another project, Swarm Forge, is intended to experiment with AI-enabled capabilities and discover new ways of fighting with and against them.

I recently wrote about what that means for a commander who has seconds to decide after machines have assembled the picture, ranked the threats, and narrowed the options. A human may remain “in the loop” and still exercise little independent judgment. Human authorization is not the same thing as human judgment, and the gap between them widens as machines take on more of the thinking.

The Pentagon already requires autonomous weapon systems to permit “appropriate levels of human judgment over the use of force.” The directive never defines how much judgment is appropriate once a machine has already shaped what the commander knows.

For most of military history, technology extended man’s physical capabilities. Gunpowder increased destructive power. Aircraft conquered distance. Radar extended perception. Computers accelerated calculation.

Artificial intelligence reaches farther. It now performs analysis, interpretation, prediction, and recommendation, cognitive work once reserved for commanders and their staffs. As machines take on that work at a speed and scale no staff can match, the officer’s role shifts from producing analysis to questioning, validating, and judging what the machine produces.

In a forthcoming address to the MacArthur Society, a nonprofit organization of West Point graduates dedicated to preserving the Academy’s traditions of “Duty, Honor, Country,” I put it this way: Speed is not wisdom, prediction is not judgment, and calculation is not conscience. The profession of arms has always required something more than selecting the most efficient course of action. Commanders must weigh what is lawful, necessary, proportionate, and ultimately right. Those judgments, and the accountability for them, belong to men and women in uniform.

That distinction also has consequences for how America determines when and how force may be used. In June, responding to Pope Leo XIV’s declaration in his AI encyclical “Magnifica Humanitas” that Just War theory is “outdated,” I argued the opposite: AI makes the tradition’s disciplines more important. Machine speed does not eliminate legitimate authority. Computational optimization does not reduce proportionality to an equation. Greater autonomy does not erase the obligation to distinguish legitimate military targets from civilians.

Scripture supplies a warning every commander should heed: “The horse is made ready for the day of battle, but the victory belongs to the LORD” (Proverbs 21:31, ESV). Judah once looked to Egypt’s horses and chariots rather than to the Lord. The temptation in every age is to trust the newest instrument of war rather than cultivate the wisdom required to use it.

The most immediate danger, then, is not the science fiction scenario of a machine rebelling against its creators. The quieter danger is that commanders grow so dependent upon machine recommendations that questioning them becomes difficult. Confidence becomes dependence, and dependence becomes deference.

Congress cannot discharge its responsibility for this transformation by funding it. It must govern it.

Lawmakers have begun confronting the problem. Congressional proposals would strengthen requirements for human responsibility over autonomous weapons, including identifying accountable commanders, preserving human judgment and control over the use of force, and maintaining records sufficient to reconstruct decisions afterward. That is an important beginning. But proposals are not guardrails until they become enforceable law.

Quantico makes that work urgent. “We need institutional knowledge and career paths for a new generation of autonomous warfighters,” Hegseth said. Once careers, budgets, and doctrine depend on autonomy, the incentives inside the Pentagon will favor more of it. Guardrails belong in place while the command is being built, not after dependence upon it becomes entrenched.

Congress should therefore turn those proposals into law, define meaningful human control, identify which decisions must remain exclusively human, and name who answers when the technology fails.

None of this requires slowing innovation to a crawl. Hegseth is right about the danger of bureaucratic inertia, and America’s adversaries give us little room for complacency. But neither technological enthusiasm nor fear of China should substitute for discernment.

America has repeatedly transformed its military while retaining something more important than any particular weapon: civilian authority, constitutional accountability, the law of armed conflict, and a professional ethic holding human commanders responsible for the forces entrusted to them.

AUTOWARCOM may ultimately give American forces an extraordinary advantage. But capability without adequate control creates strategic risk of its own.

America has already entered the age of autonomous warfare. The test now is whether our doctrine, training, congressional oversight, and moral guardrails can develop quickly enough that machines take on more of the work of warfare without assuming the moral judgment and responsibility that belong to human beings. That may be the most important command decision of the AI age.