By: Tuva Siegel, Warrior Editorial Fellow

Using AI enabled “data minutes” to support streamlined multi-domain combat is one way the U.S. The Army is starting to unlock AI to assist in “day-to-day” tasks, explains General Ronald Clark, Commander of U.S. Army Pacific. 

The use of generative AI is already embedded in the creation and streamlining of the Multi-Domain Pacific Command. Every commander’s update starts with a “data minute,” where, as Clark explains, they “spend time digging into use cases associated with generative or agentic AI.” The increasing autonomy of AI, both as a technological question and an ethical question, has long lingered throughout the recent emergence of precision drone targeting. The focus has shifted from whether AI is a “good” asset to AI as a necessary asset, but with questions of oversight. The prevailing Pentagon and Army consensus is that the technical merits with AI must be balanced alongside the primacy of human decision making. This appears to be a key focus with “data minutes” as the Army moves quickly to integrate generative AI. 

Man-Machine Interface

The Army’s approach with “data minutes” seems two-fold; the Army wants to accelerate AI integration while also ensuring human judgement and decision making remains primary. Human cognition is irreplaceable and measures are being taken to ensure the application of AI is in alignment with service’s doctrinal priorities and necessary guardrails. 

The June 18 ceremony for the redesignation of the 7th infantry defense to the Multi-Domain Command–Pacific combined a traditional ground combat force with long-range precision fires, cyber capabilities, space assets, and electronic warfare. This signaled a major strategic shift as talks around AI-enabled targeting continue to rise to a major topic of discussion among Army leaders and weapons developers. . 

This shift toward greater autonomy and increased use of AI, under human supervision, is very much grounded in the belief that soldiers are the key enabler when it comes to ultimate command and control and decision-making authority. Human leadership remains  increasingly important as the service greatly increases AI, integration. Emil Michael, undersecretary of war for research and engineering and War Department chief technology officer, outlined that AI use increased by roughly 1,420,000 users, a 1,775% increase over the past calendar year. Michael explains that the Department of War uses AI across three layered dimensions, “ the enterprise level, the intelligence level, and the warfighting level, which he described as the most important of the three.” 

AI for Patriotism 

In an effort to compete with the private sector to recruit “top-tier-talent” for rapidly advancing AI, the department is working on ways to “create sort of a … more patriotic point of view, or option, for kids coming out of college or out of grad school to come do something for their country,” Explains Michael. While the private sector's economic power can't be squandered, the department can “appeal to the next generation's sense of patriotism.” In fact,  several hundred recent graduates have already been hired as of May 20, 2026. Tara Murphy Dougherty, CEO of a prominent defense acquisition software company, further explains the high standard maintained throughout the selection process, “We are here to build a business that serves the warfighter. If that is not a mission you are interested in supporting, call somebody else,” Murphy said. 

Human-in-the-Loop

An article by Eric Jensen, former Special Counsel to the Department of Defense General Counsel, delineates the human and AI relationship into three major concepts; they are humans in-the-loop, where the machine can't make decisions without human input, on-the-loop, where the machine can't make decisions without a human overseeing the machines decision with the ability to intercept, or finally, out-of-the-loop meaning the human is not a part of the decision making process and the machine acts alone.

 Jensen explains the divisiveness of these possibilities, stating that “ the Red Cross contends that human input regarding lethal decisions is required by the law of armed conflict. Under this interpretation, the legal standard requires, at the least, a human in the loop for weapon systems that use artificial intelligence or machine learning (ML) and have the capacity to operate autonomously.” However, Jensen also emphasized the likelihood of states removing humans from the loop should the state's survival depend on it, perhaps referring to the ongoing discussion about autonomous non-lethal weapons primarily used for defensive purposes such as missile intercept. .

If a human needs to intervene during an autonomous weapons decision-making, the proper structure of authority is critical, which, for Jensen, means maintaining the ability for a human decision maker to always be able to reinsert themselves. However, even this security measure comes with its own dangers: “What circumstances warrant holding the human accountable when a decision to intervene is unfounded?” The situation is further complicated just by humans' own ability to doubt and fear their own judgment: “humans may be hesitant to override the automated system’s decisions,” solely because they don't trust their “facts” against a machine's, explains Jensen. However, there is still a paradoxical quality, as AI also has certain reliability challenges.

The DOD Directive 3000.09: Autonomy In Weapon Systems established policy and assigned responsibilities for “using autonomous and semiautonomous functions in weapon systems… to minimize the probability and consequences of failures in autonomous and semi-autonomous weapon systems that could lead to unintended engagements.” The Directive states that the policy framework for ensuring autonomous and semi-autonomous weapon systems must operate as intended across realistic combat environments, completing engagements within the boundaries of commander and operator intentions or pausing to seek human input when they cannot. The Directive mandates that systems be built with robust safety mechanisms, cybersecurity protections, and human-machine interfaces that are transparent, auditable, and understandable to trained operators. Once again, in keeping with the Army’s doctrinal focus upon human primacy, the Directive emphasizes the essential and prioritized role human judgement needs to perform. 

AI-Capable Drone Defense

For example, in the case of drone defense, human-AI teaming can enable countermeasures to be paired with targets by advanced AI enabled algorithms. Section 4 of the directive outlines the necessity of senior approval before formal development and again before fielding if: a weapon system has not previously been reviewed, including modified non-autonomous systems newly equipped with autonomous functions, or if an already-approved system has undergone substantial changes to its algorithms, mission sets, operational environments, target sets, or expected adversarial countermeasures that fall outside the scope of its original approval.

Lasers are one of several weapons used to defend against drone attacks. These countermeasures can be paired by feeding information into an AI data base, then using that AI-driven analytics to analyze a host of variables in relation to one another. All of this, as intended with “data minutes,” is to align directly to the importance of Commander decision-making.

When it comes to integrating AI into missions such as drone defense, the intent is to optimize the most effective blend between high-speed computing, advanced algorithms and human decision-making. For example, in a news story by the Navy, writer Dan Linehan illustrates what this transition to elements of AI-empowered autonomy in terms of operation actually looks like during engagement with a hostile drone. Before the integration of AI, radar would make the initial detection, followed by contact information fed over to the laser weapon system (LWS). The operator of the LWS then “uses its infrared sensor, which has a wide field of view, to start tracking the drone. Next, the high magnification and narrow field of view of its high energy laser (HEL) telescope continues the tracking as its fast-steering mirrors maintain the lock on the drone.” 

Prior to the arrival of AI-capable drone-defenses, the LWS operator would then begin the process of deciphering the drone type from a distance based off of a target reference along with “the drone’s pose, or relative orientation to the LWS, necessary for locating its aimpoints.” Then, finally, the laser beam is fired. This process is both challenging and time-consuming to conduct, with many adversities contending against the process, such as bad weather. Even the laser fire is not as simple as it is often perceived to be. Lasers destroy drones by sustained heating rather than instant destruction, requiring continuous aim at a single point, a process called aimpoint maintenance, since any movement of the drone or drift of the beam spreads the energy across a wider surface, reducing its effectiveness. This means that lasers can be scaled to disable or fully destroy/incinerate a target. 

AI & Drone Countermeasures

However, as Linehan explains, transitioning this process over to AI has been aided by the development of two major databases: one synthetic set of 100,000 images generated by Lockheed Martin and one real-world set of 77,077 images captured using a miniature Reaper drone model. Professor Brij Agrawal, Naval Post-Graduate School Department of Mechanical and Aerospace Engineering, cautioned that "if we train on only clean pictures, it won't work," noting that data must account for "different backgrounds, intensities of the sun, turbulence, and more." The AI model trained exclusively on real-world data ultimately performed best and addressed pose ambiguity. The model has since been transferred to Naval Surface Warfare Center Dahlgren for field testing, where Montag, an imaging scientist at Dahlgren, confirmed that "we now have the model running in real-time inside of our tracking system."

The transition to AI-enabled autonomy is one of many difficult questions around the role of human understanding within a process growing ever more tempted by the rapid and free-from-doubt decision-making machines. However, the importance of a human's ability to intercept, as well as the Department of War’s growing efforts to inspire the next generation to learn these systems, upholds the idea that humans should always play a role in the use of lethal force. As stated by General Clark, “it’s not about the capabilities of the equipment; it’s about the people who employ it. As a people-centric service, the United States Army puts equipment on people—we don’t put people on equipment.”

Tuva Siegel is an Editorial Fellow at Warrior Maven. She studies English at Kenyon College. Tuva is the author of Drömland, a fictional collection of short stories, and is currently studying weapons and military technology.