Ultimate Human Responsibility in Defense Systems Act of 2026

Key Points

• The bill requires the Department of Defense to ensure that all AI-powered weapons allow human commanders and operators to supervise, intervene in, or stop the use of force. This means a person must always be able to understand what the weapon is doing and shut it down if needed.
• Before any autonomous weapon capable of selecting and engaging targets on its own can be deployed, it must pass an independent adversarial "red-team" test. Experts will try to hack, jam, or trick the system under realistic combat conditions to make sure it stays safe and reliable.
• The military would create a centralized incident repository to track every time an AI weapon fails, behaves unexpectedly, or nearly causes an unintended engagement. The system is modeled after aviation safety reporting to help identify patterns and prevent future problems.
• A new Civilian Harm Mitigation and Response Office would be created within the Office of the Secretary of Defense to prevent, investigate, and respond to civilian casualties from military operations, including those involving AI weapons.
• The bill creates specialized training pipelines for military personnel working with AI weapons, including certification in human-machine teaming and mandatory training for scenarios where GPS, communications, or other systems are disrupted by enemy action.
• Within 18 months of enactment, the Director of Operational Test and Evaluation must publish unclassified testing standards for AI components in autonomous weapons, covering reliability, robustness, security, and human-machine team performance.