Trump is building killing machines and calling it innovation.
The memo that landed Friday — signed June 5, addressed to the secretaries of defense and homeland security, the attorney general, and the director of national intelligence — directs the forced integration of artificial intelligence into national security operations, from surveillance to targeting. It cites “oversight” and “civil liberties,” as if you can launch autonomous weapons that select and engage human beings without direct human control and then contain the consequences with a filing system. This is the permanent-war apparatus doing what it has done since Eisenhower named it sixty-five years ago: grabbing the next technology before anyone can ask whether it should exist, then wrapping the grab in the language of restraint. Eisenhower saw the feedback loop in 1961: an arms industry feeding a standing army, each justifying the other’s expansion. Its total influence — economic, political, even spiritual — would be felt in every city, every statehouse, every office of the federal government. He could see it then; anyone watching the Pentagon now can see the same logic applied to artificial intelligence, faster and with less friction than the nuclear arms race ever managed.
The acceleration is not a response to new threats; it is the continuation of the military-industrial complex operating at a higher volume. Andrew Bacevich documented the pattern across administrations of both parties: the “Washington rules” that demand global military presence, permanent readiness, and a procurement machine that must be fed. Autonomous weapons are the procurement machine’s dream — systems sold as precision instruments right up until they kill the wrong people, at which point their operators will call the deaths “collateral” and their manufacturers will call them a justification for the next upgrade. Removes the friction of human judgment from lethal decision-making, creates the illusion of precision while the scope of permissible conflict expands, because when the cost of initiating strikes drops to the execution of code, the political restraint that once checked American intervention evaporates. The administration trades strategic patience for algorithmic speed, mistaking the capability to kill faster for the ability to end wars, as if a faster guillotine ever ended a revolution.
A machine that selects its own targets strips the human of the moral weight that just-war doctrine — from Augustine to Michael Walzer — placed at the center of legitimate killing. Walzer’s doctrine of double effect requires that the soldier intend to strike only combatants: civilian deaths must be foreseen but not intended. An autonomous system does not intend; it processes. It collapses the distinction between combatant and civilian into probability scores. The discrimination criterion collapses, and with it the entire moral framework. What follows is not a risk of war crimes but a guarantee of systemic atrocity. When the administration shelved a broader AI executive order in May because it might slow the U.S. tech lead, it tipped its hand: the imperative is speed, not safety. The safeguards language in Friday’s memo is the residue of a debate the human-rights crowd already lost — a receipt for a purchase, not a brake on it.
The military leaders who recently urged caution on battlefield AI know what the technology does to the chain of command they claim to respect. Their caution, however, is calibrated to shape the specifications of the purchase, not to halt the expansion. The illusion that a presidential memo can maintain “oversight over autonomous weapon systems” while ordering their accelerated adoption is the illusion the MIC has always depended on: that you can uncork the killing and then control where it spills.
The soldiers tasked with executing these decisions — our veterans, those of us who served — are left carrying the moral injury of following automated orders while the architects of the war are shielded from the consequences of their algorithms. War is conducted by human beings under conditions of terror and confusion, not by sterile equations. Automating the killing institutionalizes the machinery that inflicts that injury and externalizes the risk onto the populations near the strike zone. Veterans returning to civilian life deal with the physical and psychological toll of combat; building machines that kill without human judgment does not heal those wounds.
The constitutional remedy must be immediate, because the machinery, once built, is not easily dismantled. The constitutional posture requires the preservation of human judgment in the delegation of lethal authority. Congress must revoke the open-ended authorizations that allow the executive branch to stretch the legal justification for algorithmic warfare and reassert the authority over the declaration of war. The acceleration of autonomous systems must be halted until explicit statutory limits constrain the use of weapons that select and engage targets without direct human intervention. Trump is building killing machines. The dead will be the evidence, and the oversight memos will be the epitaph — the memo is merely the receipt.