Pentagon urged to embed human authority in AI weapon rules before deadline
The administration gave the Pentagon a 90-day window, ending this week, to revise its autonomy directive so that human command remains enforceable in AI-driven weapons.
President Donald Trump ordered the Pentagon to finish a revision of its autonomy policy within 90 days, a deadline that expires this week. The memo emphasizes accelerating military AI while keeping the chain of command intact, not just adopting new technologies faster. Analysts recommend broadening the definition of autonomous weapons to cover the software that fuses sensor data and selects targets, yet argue that definition alone is insufficient.
The updated Directive 3000.09 should delineate a mission-specific authority envelope, enforce human-reserved decisions before execution, and retain detailed logs to prove human involvement. These steps align with the War Department’s AI ethics framework, which calls for traceable, reliable, and governable systems capable of being disengaged if they behave unexpectedly. By embedding enforceable human authority, the Pentagon can maintain accountability despite the rapid tempo of machine-driven decisions.
Why it matters
Ensuring human control over AI weapons is crucial to prevent unchecked autonomous lethal actions.
In this story
