WeVote
Log In

Bill

Bill

S 4682

Ultimate Human Responsibility in Defense Systems Act of 2026

119th Congress Introduced by Mark Kelly

Requires ultimate human responsibility in autonomous weapons, with design, testing, oversight, reporting, and civilian-harm safeguards to ensure human control.

Introduced in Senate
0
WeVote Research Nonpartisan
Bill Summary · S 4682

Overview

S.4682, the Ultimate Human Responsibility in Defense Systems Act of 2026, aims to ensure that human judgment remains central in the use of force by autonomous weapon systems and AI used by the Department of Defense. The bill requires design, testing, oversight, and reporting measures to guarantee “ultimate human responsibility” over any autonomously operating weapons, subject to applicable law and military policy.

Main purpose and intent

  • Establish that humans must retain ultimate responsibility for the use of force when autonomous weapon systems are involved.
  • Create a framework of design, testing, training, certification, and accountability to ensure human supervision, intervention options, and legal/ethical compliance.
  • Improve transparency and Congress oversight through annual/semiannual reporting and a centralized incident repository.
  • Address civilian harm considerations through a dedicated Civilian Harm Mitigation and Response Office.

Key provisions and changes

Definitions (Sec. 2)

  • Autonomous weapon system: a weapon that can select and engage targets without further human intervention after activation.
  • Ultimate human responsibility: a human commander’s or operator’s ability to understand context, supervise/intervene/terminate use of force, and ensure compliance with U.S. and international law.
  • Use of force: both kinetic lethal and non-lethal applications.
  • Other terms: Congressional defense committees defined per statute.

Policy and design requirements (Sec. 3)

  • The Secretary of Defense must ensure:
    • Autonomy-capable systems enable ultimate human responsibility and adhere to ethical and legal principles.
    • Development, fielding, and use align with legal frameworks and prevent unlawful AI uses.
  • Design requirements to maximum extent practicable:
    • Human supervision during mission execution where appropriate.
    • Intervention/termination mechanisms for commanders/operators.
    • Fail-safes for manual control during degradation, jamming, spoofing, or adversarial conditions.
    • Transparency in status and decision logic to aid oversight.
    • Records of target selection, decision logic, and operator actions for post-engagement review.
    • Compliance with U.S. law, international law, ROE, and treaties.

Operational oversight (Sec. 3)

  • Rigorous training for commanders/operators on supervising autonomous systems.
  • Proficiency assessments in manual target identification and threat assessment.
  • Mechanism for reporting reliability or interface concerns without fear of reprisal.

Review and certification (Sec. 4)

  • Senior-level review before fielding any autonomous weapon system capable of independent target engagement without supervision.
    • Independent red-team testing under adversarial, degraded, and contested conditions.
  • Certification by the Director of Operational Test and Evaluation that:
    • The system enables ultimate human responsibility.
    • It meets reliability and safety standards, with retesting after significant software updates or environment changes.
  • Testing standards to be published within 18 months on a publicly accessible site (classified-forms to the extent practicable).

Reporting to Congress (Sec. 5)

  • Annual starting one year after enactment and semiannual for six years:
    • Summaries of autonomous systems in development/fielded.
    • Steps taken to ensure compliance with the Act.
    • Legislative or administrative recommendations on human judgment in autonomous weapons.

Autonomous weapon system incident repository (Sec. 6)

  • Create a centralized incident reporting repository for:
    • Operational malfunctions, unintended behaviors, near-misses.
    • Incidents where target selection or engagement contradicted intent/ROE.
    • Human-machine interface failures.
    • Other safety-related events.
  • Model reporting on aviation safety systems to the extent practicable; ensure analysis, dissemination of lessons learned, and protection of sensitive information.
  • Classified semiannual Congressional briefings on trends and corrective actions.

Training pipelines (Sec. 7)

  • Develop formal training pipelines for operators, commanders, and supervisors.
  • Include certification in human-machine teaming, not just system operation.
  • Mandatory training for denied/contested environments.
  • Integrate autonomous systems into Combat Training Center rotations.
  • Continuous retraining tied to software updates and changes.

Civilian harm mitigation and response (Sec. 8)

  • Establish the Civilian Harm Mitigation and Response Office within the Office of the Secretary of Defense.
  • Responsibilities include policy development, integration of civilian-harm practices into planning, assessment/cooperation on incidents, training, and annual reporting to Congress.

Construction and scope (Sec. 9)

  • Clarifies that nothing in the Act prohibits intercepting threats with human-supervised autonomous systems.

Who/what is affected

  • Department of Defense programs and contractors developing autonomous weapons and AI capabilities.
  • Military commanders, operators, and system supervisors who supervise or interact with autonomous systems.
  • The Director of Operational Test and Evaluation (DOT&E) responsible for reviews/certifications.
  • The Civilian Harm Mitigation and Response Office within DoD.
  • Congressional defense committees receiving annual and semiannual reports.
  • The broader defense and policy community through mandated reporting and public/classified testing standards.

Procedural and timeline aspects

  • Introduction date: June 4, 2026; referred to the Senate Armed Services Committee.
  • Publicly available testing standards: to be published within 18 months after enactment.
  • Certification prerequisite: fielding requires DOT&E certification that ultimate human responsibility is maintained and that testing meets reliability/safety standards, with retesting after major updates.
  • Reporting cadence: starting one year after enactment, then semiannual for six years.
  • Incident repository: ongoing establishment and operation with required reporting and analysis.

Note: The bill establishes a robust framework emphasizing human control, safety, accountability, and civilian harm mitigation for autonomous defense systems. It requires concrete design features, oversight, testing, and transparent reporting to Congress.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.