Hegseth Promotes AI as the Future of the Military — AI Says His Orders Are Illegal

The Pentagon’s new artificial intelligence system, GenAI.mil, has declared that a hypothetical U.S. military “double strike” on a drug-smuggling boat and its survivors would be “unambiguously illegal.” The scenario presented to the AI closely mirrors real boat strike operations carried out in the Caribbean under Defense Secretary Pete Hegseth, which have resulted in at least 87 deaths and prompted bipartisan scrutiny from Congress. One of the most controversial incidents, a September 2 strike, involved a follow-up attack on two survivors who were clinging to debris in the water.

GenAI.mil, launched by the Department of Defense as an AI assistant for service members, is being heavily promoted by Hegseth, including through posters modeled after the iconic “I Want You” recruitment image — but featuring Hegseth himself. In the exchange, a user asked the AI whether ordering a second missile strike against shipwrecked survivors would violate U.S. military policy. The AI responded that such an order would clearly break the laws of armed conflict and must be refused.

The timing of the AI’s statement is notable. The Senate Armed Services Committee recently announced a bipartisan investigation into the U.S. boat strikes, with particular focus on whether the September 2 operation violated the U.S. Law of War Manual’s protections for hors de combat — individuals who are incapacitated or otherwise no longer participating in hostilities. Hegseth has denied issuing the follow-up strike order but has confirmed that the operation took place.

The controversy comes amid broader concerns about weakened legal oversight in the Department of Defense. In February 2025, President Trump and Hegseth removed several JAG lawyers, describing them as “roadblocks.” Legal analysts say the move reduced internal checks designed to prevent unlawful military actions. At the same time, the administration is advancing an Executive Order aimed at blocking states from regulating AI, even as it urges service members to adopt the Pentagon’s own system.

GenAI.mil’s assessment reinforces what military law experts and former commanders have repeatedly emphasized: targeting shipwrecked survivors is unlawful under both U.S. policy and international law. The fact that the Pentagon’s own AI tool reached the same conclusion now raises deeper questions about how closely the military’s actual operations align with the very standards it is training service members to follow.