The U.S. Air Force says its use of artificial intelligence in warfare must remain bounded by its values. Lt. Gen. Richard G. Moore Jr. has also warned that the same limits may not guide potential adversaries, leaving the Pentagon to consider how to respond if another party accepts greater autonomy in weapons systems.
Moore’s account of the Air Force’s limits
At an event at the Hudson Institute, Moore was asked how the Pentagon views autonomous warfare. He said AI ethics is discussed at the highest levels of decision-making in the Defense Department. Moore is a three-star Air Force general and deputy chief of staff for Air Force plans and programs.
In a statement to the Washington Post, Moore clarified the point he intended to make: the Air Force would not allow AI to take actions, or take actions based on information supplied by AI, unless it could ensure that the information was consistent with its values. That description places a condition on both direct machine action and decisions informed by machine output.
Moore framed that condition as an ethical commitment, while cautioning that it may not be shared by others. He said the future role of AI in warfare could depend on which societies follow the rules of warfare and which do not. He did not identify a specific potential adversary.
Autonomy raises a question of control
The concern is not only about the values used to guide a system. Giving an algorithm control over a weapon could also change who retains authority over what the system does. Mark Metcalf, a lecturer at the University of Virginia and retired U.S. naval officer, described a worst-case outcome as the party losing control after handing that control to an algorithm.
That concern offers a reason for restraint that can apply beyond differences in ethical foundations. The article points to the possibility that both the United States and China may have reasons not to give AI systems ultimate control over weapons systems. Different origins for ethical rules do not necessarily remove a shared interest in keeping control.
For military decision-makers, the distinction matters: using AI to inform a decision is not the same as allowing a system to decide and act on its own. Moore’s statement says the Air Force intends to keep its use inside a values-based boundary, even as the wider debate considers how much autonomy a weapon should have.
The pressure of a potential adversary
Moore’s remarks also point to a strategic tension. If one side limits AI action while another does not, restraint could carry a cost in a conflict, including a possible disadvantage in response speed. The source presents this as an interpretation of his comments, not as a claim that the Air Force has abandoned its stated limits.
China is described as a country the United States sees as a potential adversary in an armed conflict where AI could play a central role. Moore did not name China in his remarks, and the article does not establish that China would choose autonomous warfare. It raises the broader challenge of preparing for an opponent that might operate by different rules.
The Pentagon therefore faces two linked questions: how to preserve human and institutional control over AI-supported weapons, and how to respond if a potential adversary accepts fewer constraints. The source suggests that successfully addressing the second question could make it more likely that the principles behind the first endure over time, particularly in an actual conflict.
Ethics and military effectiveness
Keeping AI from taking action without assurance that its use fits the Air Force’s values may be a moral imperative. At the same time, the possibility of slower responses compared with a more autonomous opponent creates operational pressure. The debate is about how to hold to an ethical boundary while accounting for that pressure.
Moore’s position, as described in his statement, does not settle how the Air Force would handle every situation or what safeguards it would use. It does set out a governing principle: AI should not be allowed to act, or drive action through its information, unless its use can be shown to accord with the service’s values. Whether those principles can prevail in a conflict with a less constrained opponent remains an open strategic challenge.