Artificial intelligence is already changing military operations, from keeping equipment running to combining information from many sensors. Its role in combat is more limited: people still carry out most fighting, and the weapons described in the source generally depend on humans to choose missions, targets or clearances. The question is how far that arrangement can shift as autonomous systems develop.
AI is already part of military operations
The military uses or tests AI in four broad areas: logistics, reconnaissance, cyberspace and warfare. In logistics, AI can help optimize supply chains and predict when maintenance is needed. In cyberspace, it can help identify software vulnerabilities. Across operations, it can bring large volumes of data together to support decisions.
These applications can affect how forces prepare and respond without putting AI in charge of the fighting itself. That distinction matters: a system that sorts information or flags a problem is different from a weapon that selects and attacks a target. The technologies sit along a spectrum of autonomy, with human involvement varying by system.
Weapons differ in how much autonomy they have
One way to understand autonomous weapons is to ask where the person remains involved. Semi-autonomous systems act on a target a human selected in advance. A “fire and forget” missile, for example, can attack that designated target after a person launches it.
Human-supervised systems can identify and engage targets after activation, while a person monitors their operation. The source describes the Aegis combat system on Navy ships, as well as Patriot and Iron Dome missile defense systems, as examples. These systems are used where events can move too quickly for a person to react unaided.
Loitering munitions make the boundaries less straightforward. These aerial drones carry warheads and can be programmed by a person to attack specific targets; the person can abort the attack. They can support troops without exposing fighter planes or helicopters to the same risks. The Israeli Harop, the American Switchblade, the Russian Lancet and the Iranian Shahed are cited as widely used examples.
Their use in the Armenia–Azerbaijan clashes and the Ukraine war has prompted some military experts to view autonomy as part of deterrence. Admiral Lee Hsi-ming, former Taiwanese chief of general staff, former vice minister of national defense, and commander of the Taiwanese navy, sees loitering munitions as elemental to Taiwan's ability to deter a possible Chinese war of conquest.
Full autonomy remains a technical and human-control question
No military officially operates fully autonomous weapons systems to date, according to the source. It says the technology required for their widespread deployment does not yet exist. Although AI research has advanced, current systems struggle with context, changing circumstances and vulnerability to attack. The source also says they are not suited to making ethical life-and-death decisions.
That gap complicates the promise of speed and precision. A system may work under particular conditions yet fail when circumstances change. In a military setting, the consequences of a poor decision can be severe, so reliability is tied directly to the question of who remains responsible for using force.
There is also no simple definition of human control. Is a person in control if they launch a system that then attacks on its own? Must they be able to stop it? And what happens when events move faster than human decision-making? These questions sit at the center of debates over lethal autonomous weapons systems, often shortened to LAWS.
International efforts seek to regulate or ban such weapons, while major military powers are reluctant to abandon systems they believe could shape the balance of power. The source compares giving up advanced AI in weapons to giving up electricity and internal combustion engines, a view attributed to Paul Scharre, a former soldier, consultant to the U.S. Department of Defense, and author of “Army of None: Autonomous Weapons and the Future of War”.
Many programs still pair machines with people
Current military development focuses heavily on combining sensor data and building systems that cooperate with people. The source says some machines can fly, drive or dive, gather intelligence, attack designated targets or deliver supplies, while receiving their missions, targets and clearances from a human.
The U.S. Air Force tested variants of Kratos' XQ-58A as part of the Skyborg Program. These drones were intended to fly alongside a human pilot and provide reconnaissance and weapons support. The program has been classified since 2021, and up to a dozen drones were expected to be operational by spring 2023. The U.S. Navy is also developing autonomous tanker aircraft based on the MQ-25A Stingray drone.
Other programs described include Boeing's Loyal Wingman, sold to the Australian Air Force (RUAF); Russia's S-70 Okhotnik; and China's FH-97A. In the envisioned arrangement, a human pilot in a Next-Generation Fighter (NGF) controls accompanying drones and receives support from an AI co-pilot. European plans include a remotely controlled carrier under the Next Generation Weapon System program and the Tempest program, funded by England, Italy and Japan.
These efforts show how military AI can expand a person's ability to sense, coordinate and act. They also leave the central issue unresolved: as systems gain more freedom to act, where should human judgment and the ability to intervene remain?