Democratic senators seek AI limits in defense bill

Democratic senators are seeking to add restrictions on military artificial intelligence use to the annual defense authorization bill. The House Armed Services Committee is set to debate the legislation on Thursday. Sen. Kirsten Gillibrand introduced a bill Tuesday that would limit AI applications for launching nuclear weapons, surveilling Americans, and developing or deploying autonomous weapons.

A congressional aide said she intends to offer parts of the measure as an amendment to the Senate version of the National Defense Authorization Act. The Senate is expected to begin markup of the bill next week. The proposed Secure and Accountable Military AI Act would require human involvement in decisions involving use of force, detention, or high-consequence actions.

Amendment details High-consequence actions are defined in the bill as those involving nuclear command and control, lethal targeting, domestic surveillance, and cyber operations. These actions would also require approval from a senior-level official.

Gillibrand stated that critical national security decisions must remain with humans rather than machines. She added that the Pentagon is moving toward deploying powerful AI technology without guardrails. Sen. Elissa Slotkin plans to offer her own AI Guardrails Act as an amendment, according to NOTUS.

That measure, introduced in mid-March, would bar the Pentagon from using AI to fire autonomous weapons without human oversight, launch nuclear weapons, or conduct surveillance on Americans.

Background The amendment push follows a dispute between the Pentagon and Anthropic over contract language for AI systems. Anthropic sought limits on mass domestic surveillance and autonomous lethal weapons, while the Defense Department sought broader lawful-use provisions.

The disagreement led to the Pentagon designating Anthropic a supply chain risk, prompting the company to challenge the designation in court. President Trump signed an executive order Tuesday directing agencies to strengthen defenses against advanced AI and create a voluntary framework for early government access to company models.

Defense Secretary Pete Hegseth has advocated for AI use in classified military networks. The Pentagon announced in early May that eight major AI companies agreed to deploy systems in those networks. Last week, Vice President Vance expressed concern about AI changing warfare and stated that decisions over life and death must be made by humans.