July/August 2026
Written by Liang Xiaodong
A key U.S. Senate committee recommended imposing a regulatory framework on autonomous weapons and military artificial intelligence on the military that emphasizes human judgment and ultimate human responsibility, while largely supporting the Pentagon’s adoption of lethal autonomous weapons and urging the department to “take full advantage” of these emerging technologies.
The Senate Armed Services Committee approved the new framework in the draft National Defense Authorization Act for fiscal year 2027, and the committee finished its work on the bill on June 10.
The section related to autonomous weapons and military AI requires the department to “ensure that personnel exercise appropriate levels of human judgment” and that related systems are “designed and employed in a manner that allows commanders and operators to exercise ultimate human responsibility for the use of force.”
The requirement for an “appropriate level of human judgment” was already established in Department of Defense Directive 3000.09 on Autonomy of Weapon Systems, but a clear definition is lacking. (look activityMarch 2023. )
The new framework seeks to spell out the principles by specifying that the system must enable oversight by human operators, include methods of intervention or termination, include fail-safe mechanisms to allow manual control, provide controllers with adequate surveillance data, maintain records of target selection data and logic, and operate in accordance with “international law, rules of engagement, the laws of armed conflict, applicable treaties, and Department of Defense policy.”
The proposed legislation will codify the existing review process required by the 2023 Directive, as well as provide for verification, validation, testing and evaluation criteria. The department’s Office of the Director of Operational Test and Evaluation, whose proposed budget was cut by two-thirds in this year’s Pentagon budget request, will be responsible for ensuring that targeted systems are “sufficiently robust to perform as expected and minimize failures in an adaptive, adversarial, and realistic operational environment.”
The committee’s document would also establish an incident repository to track data on system failures, unintended behavior, and near-miss events.
The autonomy and AI language would categorically prohibit the use of these technologies in “the decision to initiate the launch or detonation of a nuclear weapon,” a move that Sen. Ed Markey (Massachusetts), Rep. Ted Lieu (D-Calif.), Rep. Don Beyer (Virginia), and Rep. Ken Buck (R-Colo.) are pushing for on a bipartisan basis in 2023.
However, the Senate Armed Services Committee’s new proposal does not include a ban on the use of AI to target nuclear weapons, as proposed by Sen. Kristen Gillibrand of New York in a standalone bill introduced on June 2.
Gillibrand’s bill is one of several introduced in the past few weeks, primarily by Democratic lawmakers, to address military uses of autonomous weapons and AI. Others included separate proposals by Sen. Adam Schiff (D-Calif.), Sen. Elissa Slotkin (D-Mich.), and Sen. Mark Kelly (D-Ariz.) working with Rep. Suhas Subramanyam (D-Virginia).
As it turns out, the Senate committee’s language largely follows the bill introduced June 8 by Democratic leaders led by Ranking Member Jack Reed (D) and Senate Defense Appropriations Subcommittee Chairman Chris Coons (D-Delaware). This document was accepted by the Republican majority as the basic text for major policy legislation, and also incorporates significant portions of the Kelly Subramanyam bill.
The Senate committee’s document is based on the Coons-Leed bill and is less specific regarding protections from domestic surveillance than other proposals put forward by Democratic senators.
Elsewhere in the bill, the committee recommends requiring the Senate to create a Robotics and Autonomous Systems Command responsible for “force generation, joint training, interoperability, doctrine development, and operational employment by other combatant commands.”
Reflecting bipartisan concerns about the impact of AI deployment, another provision in the Senate bill incorporates legislation sponsored by Kelly and Sen. Tom Cotton (R-Ark.) that would direct the Department of Defense to evaluate the effects of AI use “on maintaining and maintaining essential warfighter skills.”
Otherwise, the text largely supports the Department of Defense’s efforts to expand the use of generative AI and so-called agents, or AI-powered autonomous software.
New legislation regulating autonomous weapons systems and military AI faces an uncertain path to final passage in Congress.
The simpler House version of the defense policy bill, approved by the House Armed Services Committee on June 5, directs the Pentagon to update its directives on autonomous systems and military AI, taking into account many of the issues addressed in the Senate bill, but is less specific and more deferential to the Pentagon.
The recent surge in legislative interest in autonomous weapons and military AI in the Senate reflects growing anxiety about the Trump administration’s changes to Pentagon policy regarding the use of AI.
Secretary of Defense Pete Hegseth on January 9 directed the department to adopt a new policy requiring vendors to accept that their AI models may be deployed for “any lawful use.” The order indicated that previous sectoral restrictions would be reevaluated.
In February and March, the Pentagon took unprecedented coercive action against prominent US technology company Anthropic when it refused contract changes that removed safeguards against fully autonomous weapons systems and the use of AI models in domestic mass surveillance. (look activities, April 2026. )
President Donald Trump signed the National Security Presidential Memorandum on AI in National Security Operations on June 5, which, among other things, directs the Department of Defense to rewrite Directive 3000.09, which was last updated in January 2023.
