Senator Adam Schiff (D-CA) introduced the Human Authority in Lethal Operations Act on June 8, the first comprehensive federal legislation to restrict the Pentagon’s use of autonomous AI in combat, according to Gizmodo. The bill, known as the HALO Act, would require a human commander to retain final authority over any decision involving lethal force.
The legislation covers three distinct areas. It mandates human oversight for all lethal military AI decisions, meaning no autonomous kill chains can operate without a designated human commander at the top of the chain. It bans the use of AI for mass domestic surveillance targeting First Amendment activities. And it prohibits AI involvement in certain nuclear weapons decisions, according to Crypto Briefing.
The Anthropic Catalyst
The bill arrives directly in the wake of the Pentagon-Anthropic confrontation. Earlier this year, the Department of Defense designated Anthropic a supply chain risk after the company refused to remove guardrails that prevented its AI models from being used for mass domestic surveillance and fully autonomous weapons. That designation, the first against a frontier AI lab, triggered a cascade: the Pentagon signed contracts with OpenAI, Google, Nvidia, SpaceX, Microsoft, and Amazon Web Services to fill the gap.
“The past few months have shown us that there is an urgent need for commonsense guardrails to ensure the Defense Department’s use of AI is in line with Americans’ national security and privacy priorities,” Schiff said in a press release.
The Record-Keeping Requirement
Beyond the autonomy restrictions, the HALO Act mandates detailed record-keeping of how military decisions are made and targets are selected when AI assists the process. That documentation would be subject to later review, creating an audit trail that does not currently exist for AI-assisted military operations. The bill also establishes whistleblower protections for personnel who report violations.
Precedent for Agent Governance
The HALO Act addresses military systems, but the principles it codifies apply directly to the broader autonomous agent landscape. The core question, whether AI systems can take consequential actions without a human in the loop, is the same question every enterprise deploying agentic AI is grappling with in 2026.
The legislation arrives as the industry’s governance gap widens. A Deloitte survey of 3,235 IT and business leaders found only 21% have mature governance frameworks for agentic AI. The difference between a military autonomous weapons system and an enterprise agent authorized to execute financial transactions is one of domain, not architecture. Both are software systems making consequential decisions with varying degrees of human oversight.
Whether the HALO Act can pass a Republican-controlled Congress is another question. The Trump administration has resisted AI regulation broadly, and reports suggest the White House is warming to Anthropic again following the Mythos release. But the bill establishes a legislative marker: the first time a sitting U.S. senator has proposed codifying the principle that autonomous AI agents cannot make lethal decisions without human authority.