House Armed Services Committee members split sharply during markup Saturday over language restricting artificial intelligence from making lethal force decisions without human authorization. Progressives proposed a human-in-the-loop requirement for all autonomous weapons systems funded in the fiscal 2027 National Defense Authorization Act. Defense hawks called the amendment dangerously rigid, arguing China will field fully autonomous drones regardless of U.S. self-imposed limits.

Chairman Mike Rogers tabled a compromise requiring the Pentagon to report annually on deployments where AI recommends but does not execute lethal action. The full committee votes on the package July 2 before floor consideration later in the month.

What the Amendment Would Do

Sponsors led by Representative Sara Jacobs would prohibit procurement of systems that select targets and release munitions without a qualified operator confirming each engagement. Exceptions would cover point defense against incoming missiles where human reaction times are physically impossible. The amendment mirrors principles endorsed by a 2024 State Department political declaration signed by 47 nations but never ratified as treaty law.

Opponents including Representative Adam Smith said exceptions swallow the rule in modern air defense. Smith preferred Pentagon policy directives already requiring commanders to maintain meaningful human control, documented in updated Department of Defense Directive 3000.09.

Pentagon View

Defense Department witnesses testified that current programs keep humans responsible for targeting decisions even when AI accelerates identification. They cited Reaper drone workflows and Navy Aegis engagements as models. Researchers from the Center for Strategic and International Studies noted those examples involve human confirmation but leave room for future systems with shorter loops.

Army officials disclosed a classified pilot integrating AI for counter-drone swarms on forward bases. They declined public detail but said operators retain veto authority within two-second windows.

Geopolitical Context

Committee staff briefings cited open-source assessments that Chinese military contractors tested autonomous loitering munitions in exercises near Taiwan. Hawks argued U.S. restrictions would not influence adversary behavior and could delay fielding of defensive systems protecting troops.

Arms-control advocates submitted petitions signed by 120 former generals and diplomats supporting human-in-the-loop statutory language, though several signatories served decades ago and do not reflect current uniformed leadership views.

Funding Stakes

The NDAA authorizes $895 billion in defense spending, including $4.2 billion for joint all-domain command-and-control upgrades heavy on AI integration. Contractors Lockheed Martin and Anduril lobbied against hard bans, supporting reporting requirements instead.

Senate Armed Services begins its markup July 14. Staff expect the autonomy debate to reappear in conference committee even if the House compromise survives initial floor votes.