Democratic lawmakers have introduced a series of bills that would place new guardrails on Department of Defense (DOD) use of artificial intelligence (AI) in warfare, nuclear operations, and domestic surveillance.
Under Trump administration, the department was rebranded as the Department of War.
The legislative activity comes amid broader debate over military AI governance, including two ongoing court battles between the Pentagon and AI company Anthropic that have raised questions about the use of advanced AI systems for surveillance and military operations.
Against that backdrop, Sens. Adam Schiff, D-Calif., Elissa Slotkin, D-Mich., and Kirsten Gillibrand, D-N.Y., have introduced separate bills that would impose varying levels of human oversight, accountability requirements, and operational restrictions on AI-enabled military systems.
Schiff’s Human Authority in Lethal Operations (HALO) Act, introduced on June 8, would require a designated human commander to retain ultimate authority over any use of force involving autonomous or semi-autonomous weapons systems that use AI.
The legislation would require DOD to maintain records documenting decision-making processes and target selection for post-engagement review, establish formal review procedures before developing or fielding covered weapons systems, and conduct updated reviews when AI capabilities undergo significant modifications.
The HALO Act would also codify several AI restrictions already reflected in DOD policy, including prohibitions on using AI to profile individuals based solely on constitutionally protected activities, removing humans from decisions involving nuclear weapons employment, and acquiring personal data in violation of federal privacy protections.
Those restrictions would also apply to DOD logistics, intelligence, and operational support provided to other federal agencies.
Additionally, the bill would establish whistleblower protections for personnel reporting AI misuse, safety concerns, or civil liberties violations, including confidential and anonymous reporting channels.
Gillibrand’s Secure and Accountable Military AI Act, introduced June 2, would restrict the use of AI in nuclear weapons deployment, domestic surveillance, and fully autonomous weapons systems. Her proposal would require senior DOD approval for certain “high-consequence actions” involving AI.
In March, Slotkin introduced the AI Guardrails Act, which would prohibit the DOD from using AI to authorize nuclear weapons launches and bar autonomous systems from carrying out lethal strikes without human authorization.
The proposals come as DOD expands its use of AI across military operations.
During the current U.S. conflict with Iran, AI systems are used to process large volumes of battlefield intelligence and operational data, according to Gen. Brad Cooper, commander of U.S. Central Command.
“Our warfighters are leveraging a variety of advanced AI tools,” Cooper said in a video message posted to X. “These systems help us sift through vast amounts of data in seconds so our leaders can cut through the noise and make smarter decisions faster than the enemy can react.”
Cooper said human operators remain responsible for decisions on the use of force.
DOD v. Anthropic
Questions about military AI use have also surfaced in an ongoing legal dispute between DOD and Anthropic.
The conflict began earlier this year after Anthropic objected to the use of its technology for mass surveillance of U.S. citizens and autonomous lethal weapons systems. The dispute escalated after the federal government moved to terminate certain contracts with the company and the Pentagon designated Anthropic as a supply-chain risk.
Anthropic filed two lawsuits alleging unlawful retaliation – one in California and the other in Washington, D.C.
Most recently, Defense Secretary Pete Hegseth denied an administrative request by Anthropic to reconsider its designation as a national security risk, according to a filing in the U.S. Court of Appeals for the District of Columbia Circuit.
The decision clears the path for the appeals court to rule on the dispute. A three-judge panel considering the case appears to be divided, with one judge sharply critical of the Pentagon’s position and two others expressing skepticism toward aspects of Anthropic’s claims.
The D.C. Circuit declined an earlier request from Anthropic to pause the designation. In the California case, a federal judge temporarily barred the Pentagon from applying the national security risk designation.
The two cases create a highly complex legal limbo for Anthropic while the overarching dispute plays out in the courts. If the two courts ultimately reach conflicting conclusions, the dispute could set the stage for review by the U.S. Supreme Court.