The Senate’s version of the FY2026 National Defense Authorization Act includes provisions that would explicitly prohibit artificial intelligence from autonomously launching nuclear weapons or employing lethal force without human authorization.
The amendments, driven by Democratic Senators Elissa Slotkin of Michigan and Kirsten Gillibrand of New York, represent the most concrete legislative effort yet to draw hard lines around how the Department of Defense can deploy AI in life-and-death scenarios. Both senators sit on the Senate Armed Services Committee.
What the amendments actually do
The core mandate is straightforward: no AI system gets to decide, on its own, whether to launch a nuclear weapon or kill a person. A human being must be in the decision loop for any use of lethal force and for any nuclear weapons deployment.
Slotkin’s contribution, the AI Guardrails Act (S.4113), was introduced on March 17, 2026, and aims to embed ethical principles into federal law governing military AI. Gillibrand followed in early June 2026 with the Secure and Accountable Military AI Act, which pursues similar objectives around accountability and human oversight.
These aren’t entirely new ideas. Congress passed the Block Nuclear Launch by Autonomous Artificial Intelligence Act back in 2023, and the DOD already operates under Directive 3000.09, which addresses autonomous weapons systems. The new amendments build on that foundation.
Why this is happening now
There’s also the China factor. The international competition around military AI has intensified, with adversaries investing heavily in autonomous weapons systems.
AI policy groups have been actively advocating for these safeguards, pushing lawmakers to ensure that human decision-making stays central to lethal autonomous weapons systems, commonly referred to as LAWS.
What this means for the broader tech landscape
The NDAA still needs to be reconciled with the House version before becoming law. The bipartisan nature of these AI provisions gives them a reasonable chance of surviving the process. For anyone building, investing in, or trading around AI-related assets, the direction of travel is unmistakable: more oversight, more human accountability, and more rules.