Responsible Artificial Intelligence Defense Act of 2026

Key Points

• The bill establishes a Department of Defense policy to maximize AI and autonomy while requiring rigorous human oversight. All operations must follow the law of war, applicable treaties, weapon safety rules, and rules of engagement, along with protections for the privacy and civil liberties of Americans.
• The bill flatly bans AI from making the decision to launch a nuclear weapon. It also prohibits the military from using AI to monitor, track, or target people believed to be in the United States without a warrant based on probable cause of a crime.
• Every new autonomous weapon or AI capability must go through a two-stage review process: once before entering prototyping or formal development, and again before being fielded. Both reviews are conducted by the Under Secretary of Defense for Research and Engineering.
• AI systems must be categorized into two risk levels. Level 1 systems pose little to no risk if they fail, while Level 2 systems pose moderate to high risk to human life or use data protected under the Privacy Act or HIPAA. Level 2 systems also require a privacy impact assessment.
• Verifications of autonomous weapon systems and AI capabilities are valid for three years, after which systems must undergo a new review. The bill also creates an Autonomy and Artificial Intelligence Working Group and requires annual reports to Congress through 2037.
• Human operators must always retain the ability to detect unintended consequences and have a way to intervene, disengage, or deactivate deployed systems that show unintended or illegal behavior. Lethal force by autonomous weapons without appropriate human judgment is explicitly prohibited.