The Trump Regime Just Ordered the Military to Ignore AI Safety Guardrails…

 Brent Molnar: A Voice of Reason

  ·

Follow

rpotnSeods857uf37530h5h16utmm71hti2uca14c7aa252tgcal806m6chh ·

The Trump Regime Just Ordered the Military to Ignore AI Safety Guardrails…

A company that builds artificial intelligence tried to write a contractual guarantee that its own technology would not be used to kill people without a human making the final decision. The United States government just told them no.

That is not a hypothetical. That is what happened this week when the White House released a National Security Presidential Memorandum on Artificial Intelligence, and Defense Secretary Pete Hegseth confirmed that the administration is pushing back against firms like Anthropic, which had requested contractual guardrails explicitly prohibiting the deployment of their technology in fully autonomous weapons systems. The administration's answer was, in effect, that American weapons contractors and military commanders should not be constrained by the safety preferences of the companies whose technology they intend to use to wage war.

The directive itself is sweeping. It commands the military and all national security agencies to aggressively accelerate the integration of advanced commercial frontier AI into weapons systems and logistics. It also formally rescinds the previous administration's AI governance framework, which had established baseline protections governing how AI could and could not be deployed in national security contexts. The Trump administration did not replace that framework with a stricter one. It replaced it with nothing.

To be precise about what is being described here: fully autonomous weapons systems are weapons that identify, select, and engage targets without a human being in the decision loop. A machine assesses a threat. A machine decides to fire. A machine fires. No soldier. No officer. No authorization. The question of whether a human being lives or dies is delegated entirely to an algorithm. That is what Anthropic's engineers apparently understood well enough that they tried to prohibit it by contract. That is what Pete Hegseth just said is an unacceptable restriction on American military power.

The rescission of the previous administration's framework is not a minor procedural footnote. The Biden-era AI executive order, issued in October 2023, represented the most comprehensive attempt by any American government to establish binding standards for AI safety, security, and accountability across the federal enterprise. It required agencies to conduct safety testing before deploying AI in high-risk contexts. It created disclosure requirements. It established, at minimum, the principle that humans should remain accountable for consequential AI-driven decisions. The Trump administration did not refine those standards. It erased them.

This is a pattern, not an anomaly. Since January 2025, this administration has systematically dismantled the oversight infrastructure across virtually every federal agency. The Consumer Financial Protection Bureau was gutted. The inspector general system was purged. The Department of Justice has been converted into a loyalty instrument. At every turn, the operating principle has been the same: remove the mechanisms designed to impose accountability on the exercise of power, and remove them as fast as possible. The AI memorandum is that same project, applied to the technology that may soon decide who gets targeted in a war zone.

There is something worth sitting with about the specific complaint Hegseth is making. Private companies, using private capital, built large-scale AI systems and then said: we will work with you, but not if you intend to use what we built to kill people without a human being responsible for that decision. That is not an unreasonable position. It is arguably a morally serious one. The administration's objection is that this constitutes an unacceptable constraint on military flexibility. Which means the administration is affirmatively arguing that it should retain the right to deploy autonomous lethal force free from the ethical objections of the people who built the tools being used.

Legal scholars and international humanitarian law experts have warned for years that autonomous weapons systems create an accountability vacuum that existing laws of war were never designed to address. If a machine selects and kills the wrong target, who is responsible? The programmer? The commanding officer who deployed the system? The defense contractor who sold it? The answer, under current frameworks, is genuinely unclear. The Geneva Conventions presuppose human judgment. They presuppose a person capable of distinguishing a combatant from a civilian, of assessing proportionality, of making the kind of moral calculation that international law demands before lethal force is applied. An algorithm does not make moral calculations. It makes pattern matches.

Congress has said very little. That silence is itself a statement. The National Security Presidential Memorandum does not require congressional authorization. It is an executive directive, signed by one man, reshaping the terms under which the United States military may deploy machines to make kill decisions. The constitutional framework vests the war power jointly in the legislative and executive branches precisely because the founders understood that decisions about life and death at scale should not rest in a single set of hands. That framework is not being formally violated here. It is simply being ignored, in the way that this administration has learned to ignore every structure that might slow it down.

What is being rescinded is not just a policy document. What is being rescinded is the principle that a human being must be accountable for the decision to take a human life. Every safeguard that the previous administration tried to build into AI governance, every contractual guardrail that a private company tried to negotiate, every inspector general and oversight mechanism that this administration has systematically removed since day one, all of it points in the same direction. Power without accountability. Force without authorization. Decisions without anyone to answer for them. That is not a national security strategy. That is the architecture of something else entirely. And we should call it by its proper name.