S-4707-119
Read twice and referred to the Committee on Armed Services.
Sponsored by Christopher Coons (D-DE)
What it does
This bill would amend federal law to establish a formal policy and oversight framework for how the Department of Defense (DoD) develops, tests, and deploys artificial intelligence (AI) and autonomous weapon systems. It would require two-stage review and verification by the Under Secretary of Defense for Research and Engineering before any such system enters development or is fielded. The bill would also prohibit AI from making nuclear launch decisions, ban warrantless AI-driven surveillance of people in the United States, and bar lethal autonomous weapons from using force without appropriate human judgment. It would create a standing working group, mandate annual congressional reports through 2037, and require privacy impact assessments for AI systems that handle protected personal data.
Who benefits
U.S. military personnel who would operate AI-enabled systems with clearer safety standards and override capabilities. Civilian noncombatants in conflict zones who would benefit from law-of-war compliance requirements built into weapon system design. U.S. persons whose privacy would be protected by the prohibition on warrantless AI surveillance. Defense contractors who build compliant systems and gain a clearer regulatory pathway. Congressional defense committees, which would receive annual reports and a 30-day notification window before certain exempted systems are deployed. Civil liberties organizations and the public broadly, through the privacy impact assessment requirement.
Who is hurt
Defense contractors and military program offices whose AI or autonomous weapon systems would face new multi-stage review requirements, potentially slowing acquisition timelines and increasing development costs. Operational commanders who may find that review and verification requirements reduce flexibility in rapidly evolving conflict scenarios. Smaller defense technology firms that may lack resources to navigate the new compliance framework. Adversary-focused intelligence operations that rely on autonomous or AI-driven surveillance tools, which could face operational constraints. Taxpayers who may bear the cost of expanded testing, evaluation, and oversight infrastructure within DoD.
Supporters argue
Supporters argue that as the U.S. military increasingly integrates AI into lethal decision-making, the absence of statutory guardrails creates serious risks of unintended civilian casualties, escalation, and adversary exploitation of system vulnerabilities. They contend that codifying human oversight requirements into law — rather than leaving them to changeable internal DoD directives like Directive 3000.09 — provides durable, enforceable standards that keep pace with rapidly advancing technology. They further argue that the explicit prohibition on AI-initiated nuclear launches and warrantless domestic surveillance addresses the most catastrophic potential misuses of autonomous systems before they can occur.
Opponents argue
Opponents argue that layering statutory review requirements onto military AI acquisition could create bureaucratic bottlenecks that slow deployment of systems the U.S. needs to maintain technological superiority over adversaries like China and Russia, who face no equivalent constraints. They contend that the bill's two-stage verification process and mandatory working group add procedural overhead that existing DoD directives already address more flexibly, and that rigid statutory definitions of "autonomous weapon systems" and "appropriate human judgment" may not translate well across the full spectrum of rapidly evolving AI capabilities and operational contexts.