One Step Away from Terminators

Recently, at a press convention at Ukrinform, Deputy Commander-in-Chief of the Armed Forces of Ukraine Brigadier General Andrii Lebedenko in contrast the navy innovation system to a steady conveyor belt: fight items determine an issue, navy analysts formulate the technical necessities, and builders create options that, after profitable testing, are instantly despatched to the entrance.

In the primary half of 2026 alone, the Armed Forces of Ukraine performed 433 exams of recent weapons and navy gear and evaluated almost 600 modern tasks, greater than 40 of which acquired state funding.

These figures are now not about remoted technological experiments. War is more and more entrusting machines with duties that had been beforehand carried out by people. And it isn’t nearly drones and floor robotic techniques. Artificial intelligence is step by step being utilized in reconnaissance, air protection, command and management, navigation, and logistics.

How far has this course of already gone, and the place is it main? Where does the machine assistant finish and an algorithm able to appearing autonomously start? And are we actually approaching the age of Terminators?

Ukrinform got down to discover out.

During the press convention at Ukrinform. Photo by Yuliia Ovsiannikova

FIND, IDENTIFY, DESTROY

Imagine a standard reconnaissance drone. It is flying over enemy positions and transmitting a stay video feed to the operator. In the previous, the operator needed to spend hours watching the display, looking out amongst bushes, buildings, and roads for a Russian tank or artillery piece. Now a few of that work will be handed over to a pc.

An AI algorithm analyzes the video alongside the operator and identifies targets: a tank, an armored car, an plane, or different navy gear.

DELTA. MilitaryInkind photograph

This is how Ukraine’s modern Avengers synthetic intelligence system works, built-in into the DELTA fight ecosystem. It analyzes video from drones and robotically searches for Russian navy gear. According to Ukraine’s Ministry of Defense, the system can determine an object in a video stream inside seconds. For a human operator, this implies one thing quite simple: there isn’t a want to observe your entire video—the pc robotically filters out irrelevant footage.

This is step one. But it will get extra attention-grabbing from right here.

Finding a goal doesn’t essentially imply hitting it. The operator has to manage the drone, preserve the goal in view, and alter its flight path. Meanwhile, the enemy is jamming the communications hyperlink. So the following step is to show the drone to grasp the place it’s going by itself. The Ukrainian Dovbach system, for instance, makes use of a digital camera and specialised software program that robotically identifies the goal and helps the drone keep its course towards it.

Image from the monitor of the ZIR AI (Ardupilot) computerized terminal-guidance module. Photo courtesy of the producer

There are different comparable options as nicely. For instance, ZIR AI (ZIR) is a Ukrainian synthetic intelligence and computer-vision system for unmanned plane that permits FPV drones to robotically detect, determine, and strike enemy targets with out human involvement on the closing stage. Put very merely, that is what occurs: a human reveals the machine the place to fly, and the AI then figures out by itself the way to get there.

According to Ukraine’s Ministry of Defense, Ukrainian Defense Forces items are already utilizing greater than 70 completely different techniques based mostly on synthetic intelligence and machine-vision applied sciences to interact enemy targets.

Moreover, greater than 200 Ukrainian corporations are engaged in growing AI-powered drones. On Brave1 Market, the state defense-technology market, Ukrainian defenders can select from 46 AI-related options, together with target-recognition, optical-stabilization, and computerized terminal-guidance modules.

The aim introduced by Ukraine’s Ministry of Defense is formidable: all Ukrainian drones on the entrance must be geared up with laptop imaginative and prescient and AI.

The Dovbach system: laptop imaginative and prescient for deadly precision. Photo courtesy of the producer

HOW DRONES ARE GETTING SMARTER

One of the important thing capabilities of the ZIR AI system is its means to acknowledge seven kinds of targets: infantry, civilian automobiles, vans, air-defense techniques, artillery, tanks, and armored automobiles. The neural community is skilled to detect enemy gear even when it’s camouflaged in a tree line. It may function regardless of enemy digital warfare (EW): as soon as the pilot or the AI locks onto a goal, the drone switches to totally autonomous flight. So even when an enemy electronic-warfare system utterly jams the communications hyperlink and GPS sign, the drone will fly to the goal and destroy it by itself. Its autonomous flight vary inside an EW protection space is greater than 3,000 meters.

In addition, ZIR can reliably monitor and precisely have interaction shifting targets touring at speeds of as much as 60–110 km/h. Its Circular Error Probable (CEP) is lower than one meter.

The purpose Ukrainian drones are getting “smarter” so shortly is, surprisingly, the struggle itself. More exactly, it’s the huge quantity of information generated by the struggle, above all video footage.

In August 2026, Ukraine’s Ministry of Defense opened access to the Avengers Labs laboratory for Ukrainian MilTech corporations and drone builders.

Access to Avengers Labs accelerates the event of AI-based applied sciences. Image from the web site of Ukraine’s Ministry of Defense

This huge library incorporates a singular annotated dataset of 5 million movies of actual fight operations. Access to this database permits personal drone builders (such because the aforementioned ZIR AI), who would in any other case wrestle to gather such a big quantity of high-quality materials on their very own, to coach their neural networks to acknowledge tanks, artillery, infantry, air-defense techniques, and even Shahed drones—shortly and freed from cost.

In this fashion, numerous drone fashions skilled at Avengers Labs change into the premise for mass-produced unmanned plane.

So that is already a actuality: if an operator can’t ship a command to a drone, the machine is able to considering for itself and understanding the place it’s, the place it’s going, and what it sees in entrance of it.

Among the priorities mentioned on the press convention by representatives of the Central Directorate for Innovation Activity of the Armed Forces of Ukraine and navy analysis establishments weren’t solely growing the autonomy of UAVs, growing jam-resistant navigation, and utilizing AI. They additionally mentioned the following step: instructing unmanned plane to function solely with no pilot and navigate autonomously.

AVOID GETTING LOST AND BYPASS AIR DEFENSES

This is especially necessary for strikes deep inside Russia. Since GPS alerts will be jammed alongside a route over enemy territory, a drone dangers getting misplaced with out navigation.

Ukrainian builders are at present instructing their unmanned plane to navigate actually “by the picture.” The digital camera sees roads, rivers, forests, buildings, and different distinctive landmarks, whereas the pc compares this imagery with a satellite tv for pc map saved in its reminiscence and makes use of the comparability to find out its location. In different phrases, an individual can launch the drone, and it may then alter its personal route.

This expertise is now being utilized in middle-strike techniques—drones that function at appreciable distances from the entrance line. AI additionally performs a further job right here: serving to plot a route that avoids Russian air defenses.

The drone can analyze the placement of radars, take their protection areas into consideration, and search for a much less harmful route. As it approaches the designated space, it may change machine imaginative and prescient again on. The digital camera sees objects, and the algorithm determines what they’re. If the specified goal seems amongst them—for instance, a Russian radar—the drone adjusts its flight path. In different phrases, a single plane can successfully carry out a number of phases autonomously: navigate to its vacation spot, keep away from harmful areas, find the required object, and residential in on it.

THE OPERATOR IS NO LONGER A PILOT

One of probably the most telling examples of the event of navy AI is the battle towards Russian Shahed drones.

During nighttime assaults, when dozens or a whole lot of enemy drones are within the sky, the traditional strategy requires operators to identify every enemy UAV, determine it, launch the suitable interceptor, and hit the goal. The trendy various is an automatic system that performs virtually all of this work by itself.

In June 2026, Ukraine’s Ministry of Defense introduced profitable testing of such a expertise and the beginning of its scale-up.

According to the data launched, synthetic intelligence now automates as much as 95 % of the interception cycle, eradicating the human issue from probably the most difficult phases of the hunt. The operator merely designates the goal and offers the command to shoot it down. The system then independently directs the air-defense drone to the interception space, locates the Shahed, acquires it within the digital camera’s view, tracks it, and destroys it.

At the identical time, the human doesn’t disappear from the method. They merely now not management each motion of the interceptor manually, however stay the hyperlink that makes the important thing determination. In different phrases, the operator ceases to be a standard pilot and turns into the machine’s commander.

This transformation is especially necessary now, because the Russians are ramping up manufacturing of their drones. On March 24 this yr, Russia launched the most important assault on Ukraine at that time, with almost 1,000 assault UAVs in a single day. And if the enemy, having elevated manufacturing, begins launching considerably extra UAVs at Ukraine, we merely is not going to have sufficient individuals to manually management 1000’s of interceptors. That is why the present emphasis is on producing and procuring modern interceptor drones.

Octopus-100 interceptor drone designed to destroy jet-powered Shaheds and different high-speed aerial targets. Photo courtesy of the producer

For instance, Ukraine’s Ministry of Defense just lately bought 8,000 Octopus interceptor drones particularly designed to destroy Shaheds. Dozens of Ukrainian corporations are concerned of their manufacturing. At the identical time, the JEDI Shahed Hunter, Shvidun, and different interceptors are additionally being developed. They have various levels of autonomy, and never all of them use AI to the identical extent. But the path sought by the Ukrainian navy is obvious: the extra targets there are, the much less time a human ought to should spend manually controlling every weapon system.

Shvidun interceptor drone. Photo courtesy of Ukraine’s Ministry of Defense

FROM A SIMPLE FUNCTION TO AN INTELLIGENT SWARM

There is one other necessary facet that’s simple to miss. AI in warfare is not only about unmanned plane. At the identical press convention, Andrii Lebedenko spoke individually about growing software program to course of uncooked radar information.

What does that imply in easy phrases?

A radar sees an enormous variety of objects within the sky. A human has to find out that are plane, that are missiles, that are enemy drones, and that are, for instance, interference or false targets. A pc can carry out this job a lot sooner than a human, which is very necessary when there are giant numbers of targets. So AI is step by step turning into not solely a drone’s “eyes,” but additionally an air-defense operator’s assistant.

Another benefit of such an answer is that people have bodily limitations, whereas an algorithm doesn’t. The Avengers system we talked about earlier is a very good instance. Its major benefit is just not that it sees higher than a human. The digital camera sees the identical factor. The distinction is {that a} machine doesn’t get bored with watching.

For instance, a whole lot of drones might transmit video concurrently. No operator can successfully monitor such a kaleidoscope of video feeds. A pc can.

It can filter out many of the footage and present the operator solely these segments that comprise indications of enemy gear. So AI doesn’t essentially should make selections by itself. Sometimes its most precious operate is way less complicated: stopping a human from drowning in data. What turns into data overload for a human is, furthermore, uncooked materials for AI: the extra information it has, the extra alternatives it has to study and enhance.

AI wants information. The extra real-world examples it sees, the higher it performs. And struggle generates huge quantities of such information. Every reconnaissance flight produces video. Every detection of a Russian tank offers a brand new instance. Every misguided consequence produced by an algorithm turns into data that can be utilized to enhance it.

That is why Ukraine’s Ministry of Defense is giving builders entry to actual fight information to coach new drone fashions. This creates a form of improvement suggestions loop: the entrance offers information that trains AI, the improved algorithm returns to the entrance, and the entrance in flip generates new information.

And so the cycle repeats. This is one purpose why technological warfare can evolve a lot sooner than typical warfare: every new deployment not solely produces a consequence but additionally leaves behind worthwhile reference materials for the following stage of improvement.

However, instructing an algorithm to behave autonomously is simply step one. The actual breakthrough will come when such machines can work not individually, however collectively. That is why the Armed Forces of Ukraine are speaking about swarm applied sciences.

Why may a swarm be a really revolutionary answer? The reply lies in how a number of machines can work collectively, dividing the duties amongst themselves. One drone finds the goal, one other identifies it, a 3rd assaults, and a fourth displays the airspace. If considered one of them is misplaced, the system instantly redistributes its duties amongst those who stay.

For a human, making so many choices concurrently is just about unattainable. For an algorithm, it’s exactly the form of job it was designed for. So the near-term way forward for AI in warfare might not essentially be an “intelligent” robotic that may do the whole lot directly, however teams of comparatively easy machines able to working as a single system.

“THE TERMINATOR” IS ALREADY HERE—IT JUST LOOKS DIFFERENT

Ukrainian navy officers on the press convention additionally spoke individually in regards to the improvement of Russian navy applied sciences.

Russia continues to work on missiles, however it is usually devoting growing consideration to unmanned techniques. Among its priorities are new variants of Shaheds, small drones able to working with out communications or navigation, naval drones, and defensive techniques. At the identical time, the Russians have begun integrating completely different techniques right into a single community, linking radars, electronic-warfare techniques, detection techniques, interceptor drones, and assault UAVs right into a single built-in community.

According to Brigadier General Andrii Lebedenko, Russia’s Rubicon is already utilizing the same precept: first, a goal is detected; then the data is processed, after which the suitable drone is introduced in to hold out the mission.

In different phrases, the primary competitors now’s over the velocity of decision-making. Whoever spots the goal first, identifies it, and strikes first beneficial properties the benefit. This is exactly the place synthetic intelligence can present a decisive time benefit.

The closing say ought to at all times stay with a human. Photo courtesy of Ukraine’s Ministry of Defense

Of course, the human stays within the loop. They outline the mission, choose the goal, launch the system, and, if essential, can intervene or abort the assault. But machine intelligence is step by step being given better freedom inside the parameters of the assigned job.

At the identical time, applied sciences are being developed to scale back the time between detecting a risk and destroying it. The press convention coated laser weapons, pulsed techniques for countering drone swarms, autonomous platforms, and new air-defense options. According to Lebedenko, the primary Ukrainian laser techniques are already able to participating FPV drones, whereas pulsed techniques can concurrently disrupt the electronics of a number of unmanned plane.

This is now not a prediction. Some of those applied sciences are already in use on the Ukrainian entrance, which has change into one of many world’s major proving grounds for the event of fight AI.

For Ukraine, this isn’t merely a technological competitors with the enemy. It is a matter of survival. The benefit will go to whoever has machines that may see sooner, perceive the state of affairs extra precisely, and act collectively.

But synthetic intelligence has no morality. That is why the ultimate say ought to at all times stay with a human—the one that units the target and bears duty for the choice. So we’re unlikely to see Hollywood-style Terminators on the battlefield.

The machines will likely be completely different—smarter, sooner, and extra autonomous. But for us, the true “Cyborgs” will perpetually be individuals—the defenders who held Donetsk Airport in 2014.

Ivan Stupak, Kyiv

Photos by Yuliia Ovsiannikova, courtesy of Ukraine’s Ministry of Defense, and from open sources