S-4113-119
Read twice and referred to the Committee on Armed Services.
Sponsored by Elissa Slotkin (D-MI)
What it does
This bill would prohibit the Department of Defense from using artificial intelligence to launch or detonate nuclear weapons, to monitor or target individuals or groups in the United States without an individualized legal basis, or to employ lethal force through autonomous weapon systems without appropriate human oversight. It would also create a waiver process allowing the Secretary of Defense — without delegation — to temporarily exempt specific autonomous weapon systems from the lethal-force prohibition, provided the system's error rate does not exceed that of trained human operators and Congress is notified within five days.
Who benefits
U.S. civilians and civil liberties advocates who would gain statutory protection against AI-driven domestic surveillance by the military. Members of Congress who would receive new oversight authority through mandatory waiver notifications. Advocacy groups and communities historically subject to government surveillance. International humanitarian law proponents who support human control over lethal force decisions. Domestic technology companies whose AI products would be subject to clearer legal boundaries when contracting with DoD.
Who is hurt
Defense contractors and AI developers who may face narrowed markets or added compliance requirements for autonomous weapons systems. Military commanders and operational planners who would have less flexibility to deploy autonomous systems in time-sensitive scenarios. DoD program offices that would need to redesign or re-certify existing or planned AI-enabled weapons systems. Allies and partner nations whose joint operations with U.S. forces could be complicated by unilateral U.S. restrictions on autonomous systems. Taxpayers who may bear costs of compliance, testing, and certification requirements.
Supporters argue
Supporters argue that placing AI in control of nuclear launches or lethal force without human judgment creates catastrophic and irreversible risks — a single algorithmic error could trigger mass casualties or an unintended war. They contend that the bill codifies existing DoD Directive 3000.09 into law, giving it enforceable teeth, and that the domestic surveillance prohibition directly addresses documented concerns about military AI tools being turned inward on American citizens in violation of the First and Fourth Amendments. The waiver mechanism, they argue, preserves operational flexibility for genuine national security emergencies while ensuring Congress remains informed.
Opponents argue
Opponents argue that statutory restrictions on autonomous weapons could place U.S. forces at a decisive disadvantage against adversaries — particularly China and Russia — who face no equivalent constraints and are rapidly fielding autonomous systems. They contend that the "appropriate levels of human judgment" standard is vague and could be interpreted inconsistently across programs, creating legal uncertainty that slows legitimate AI adoption. Critics also argue that the human-operator error-rate benchmark for waivers is technically difficult to measure and may be unworkable in novel operational environments with no human performance baseline.