Overview
- The Senate version of the FY2026 National Defense Authorization Act bars any AI system from autonomously launching nuclear weapons or using lethal force without a human explicitly authorizing the action.
- Democratic senators Elissa Slotkin and Kirsten Gillibrand led the push to write these human‑in‑the‑loop requirements into the defense bill and to codify accountability for lethal uses of AI.
- A related Senate amendment from Sen. Mark Kelly requires what it calls ultimate human responsibility over force decisions and would codify the Pentagon’s Directive 3000.09 into statute.
- Supporters point to technical failures and security risks, including recent high‑severity flaws found by AI firms, and to rival countries’ investment in autonomous weapons as drivers of the urgency for clear legal guardrails.
- The provisions are bipartisan and build on a 2023 law and existing DOD rules, but they are not yet law and leave open practical questions about enforcement, audits, procurement rules, and how fielded systems will prove meaningful human control.