“Right now, the Pentagon is moving toward deploying incredibly powerful AI technology without commonsense guardrails in place, which could have catastrophic consequences that make all of us less safe,” Sen. Kirsten Gillibrand said in an emailed statement.
Sen. Kirsten Gillibrand’s Secure and Accountable Military AI Act
Sen. Kirsten Gillibrand, D-N.Y., on Tuesday announced legislation titled “The Secure and Accountable Military AI Act” that would restrict the Pentagon’s use of artificial intelligence in operations and impose heavy regulation on its use in fully-autonomous weapons, for domestic surveillance, and with nuclear weapons. Gillibrand’s office said other lawmakers have not yet joined the legislation, and she plans to offer elements of the bill as amendments to the Senate’s version of the National Defense Authorization Act (NDAA).
High-consequence uses: nuclear missions, lethal targeting, domestic surveillance, cyber
The bill directs the Defense Secretary — named in the text as Pete Hegseth — to designate specific AI uses as “high consequence,” explicitly listing nuclear missions, lethal targeting, domestic surveillance, and cyber. For operations so designated, the statute would require written approval from an undersecretary or the vice chairman of the Joint Chiefs of Staff before AI use. The bill also includes a congressional notification regime: 15 days’ notice to Congressional defense committees prior to using AI for those high-consequence operations, with a 48-hour notification window allowed after deployment in certain circumstances.
Human accountability and a clear chain of command
One core aim of the bill is to establish a clear chain of human accountability for AI on the battlefield. Gillibrand’s proposal asks the Pentagon to make explicit who the accountable human decision-maker is, or what accountability chain exists, when AI-guided technology is used during high-consequence operations. The language frames that clarity as a safeguard to “keep humans in charge and keep AI’s use in warfare smart and safe,” according to the senator’s statement.
Requirements for AI contractors and frontier AI companies
The legislation would also impose reporting obligations on AI contractors. According to a summary from Gillibrand’s office, contractors would be required to rapidly report certain incidents to the Pentagon “including theft of model weights or data poisoning.” If enacted, the Defense Department would need to be notified within three days for security breaches and within seven days for “concerning model behavior.”
Becca Wasser, the defense lead for Bloomberg Economics, described the bill as formalizing longstanding norms and setting guardrails as private AI companies become more enmeshed with the Pentagon. “In some ways it's not novel, but it is codifying things in many respects that have been long-standing norms,” Wasser said, adding that the measure could act as “a check on the Pentagon's full embrace of AI and private companies.”
Responses from the White House, other senators, and industry tensions
The bill was unveiled the same day President Donald Trump signed an executive order urging that AI be “deployed rapidly to confront any and all threats” to the United States. Trump’s order said the U.S. “refuse[s] to stifle this innovation with overly burdensome regulation.”
Separately, NOTUS reported that Sen. Elise Slotkin, D-Mich., plans to insert a similar AI-guardrails bill — introduced earlier this year — into the NDAA. The Senate Armed Service Committee is slated to mark up the annual defense policy bill next week, creating a near-term opportunity for those amendments.
The proposal also comes after public friction earlier this year between the Defense Department and the AI company Anthropic, centered on the company’s objections to potential military uses of its technology. That dispute figures in the broader debate the bill seeks to address: how to reconcile rapid AI adoption with controls and contractor obligations.
What this means for the Pentagon, AI companies, and Congressional defense committees
- Pentagon: Would face new internal designation and approval requirements for high-consequence AI operations, be asked to clarify accountable human decision-makers, and absorb new contractor incident reports within specified timeframes.
- AI companies and contractors: Would be required to report incidents such as theft of model weights or data poisoning within strict windows (three days for security breaches, seven days for concerning model behavior), increasing operational transparency to the Department of Defense.
- Congressional defense committees: Would receive notification rights tied to high-consequence AI operations — 15 days before use in most cases, or 48 hours after deployment in specified circumstances — and could see Gillibrand’s provisions considered as NDAA amendments during the committee mark-up next week.
The bill frames itself as a set of “commonsense guardrails,” but it arrives at a moment of competing impulses: a presidential directive to accelerate AI deployment and lawmakers seeking statutory limits and clearer lines of human accountability. With the Senate Armed Service Committee set to mark up the NDAA next week and Gillibrand planning to press parts of her measure as amendments, the immediate question is whether those statutory guardrails will be folded into the broader defense authorization or remain separate proposals.
