General AI: How artificial intelligence became a key tool for the Ukrainian military

The war in Ukraine has produced one of the largest revolutions in military affairs in history, argues former Google head Eric Schmidt, who has since gone on to found the military startup Perennial Autonomy (previously known as Swift Beat). Schmidt’s company is one of the key players in the AI transformation that is currently underway on the Ukrainian front. Until relatively recently, tanks, aircraft carriers, and strategic bombers were seen as the main symbols of a country’s military power. Today, however, digital technology, data-processing speed, artificial intelligence, and unmanned systems are at the very center of how modern armies wage war. Ukraine has become a unique testing ground for new military approaches under real combat conditions, attracting interest not only from weapons manufacturers but also from companies working with data, software, robotics, and artificial intelligence. Among them are Alex Karp’s Palantir and Schmidt’s Perennial Autonomy. Cooperation with those companies gives Ukraine’s army access to advanced technologies, helps attract investment in defense tech, and accelerates the development of Ukraine’s own defense technology sector.

Alex Karp, CEO of the U.S. tech giant Palantir, personally visited Kyiv in June 2022, becoming one of the first major Western IT company heads to travel to the Ukrainian capital after the start of Russia’s full-scale invasion. The two sides later held repeated meetings both in Ukraine and at international venues to discuss cooperation in defense technology and artificial intelligence.

Palantir executives and employees continue to regularly visit Ukraine. The company also has a small team of local specialists on the ground in Kyiv who adapt Palantir software to front-line needs and serve as a link between the company and Ukrainian government agencies.

Founded in 2003, Palantir is one of the key players in the global defense tech sector. It specializes in analyzing large amounts of data and creating software for the government agencies, intelligence services, and militaries of many countries. In the U.S. alone, its major clients include the CIA, the FBI, the Pentagon, the Air Force, the Marine Corps, and Special Operations Command.

Palantir’s technologies are also used by the British army and Israel and have been tested by militaries in Japan, Norway, and Denmark. After the start of Russia’s full-scale invasion, Ukrainian ministries joined that list, including the Ministry of Digital Transformation, the Ministry of Education and Science, the Ministry of Economy, and the Ministry of Defense.

For Palantir Ukraine has effectively become its largest testing ground for military AI development under the conditions of modern combat.

“For us, this is the next step in developing win-win cooperation. Partners get the opportunity to train their AI models on real data from a modern war, while Ukraine gets faster development of autonomous systems and new technological solutions for the front,” then-Defense Minister Mykhailo Fedorov said after meeting Karp in May 2026.

Ukraine has effectively become the largest testing ground for military AI development under the conditions of a modern conflict

As early as 2023, Fedorov, who was then Ukraine’s Minister of Digital Transformation, described Kyiv’s joint work with Palantir to develop technology capable of analyzing large volumes of data and turning them into easily digestible information for ministries, government agencies, and the public. In 2023 and 2024, Ukraine and Palantir worked together on demining projects and initiatives aimed at the safe return to classrooms of children who had been forced into online learning because of Russian aerial attacks.

Time magazine reported in early 2024 that the Ukrainian military had begun using Palantir-developed technology to analyze vast amounts of intelligence data and plan its combat operations. In modern war, information flows through dozens of systems that often cannot share data quickly, leaving commanders waiting hours or even days to act, and Palantir’s technologies shorten that gap. The company’s platforms allow data from satellites, drones, and other surveillance systems to be processed in a single environment, enabling commanders and analysts to assess developing situations more quickly. Palantir software gives the military a more complete picture of events on a specific section of the front in near real time.

Palantir gives the Ukrainian military a more complete picture of events on the battlefield in near real time

One of Palantir’s best-known products is Gotham, a platform designed for military and intelligence agencies. The decision-making system gathers data from satellites, sensors, radar, bank transactions, and social media, then identifies links between events, objects, and people. For example, it can help spot connections between the movement of equipment, satellite images, and intelligence reports that at first glance appear unrelated.

Gotham is described as an “AI-powered kill chain,” allowing users to task sensors to quickly detect a target and select a means of striking it — either automatically or manually. In other words, it can find the proverbial “needle in a haystack,” then remove it.

An AI expert who asked to remain anonymous told The Insider that the platform looks like Palantir’s answer to the concept of network-centric warfare and its technical embodiment for a potential conflict with China, embodying the doctrine of Joint All-Domain Command and Control, as well as the principle of “any sensor, any decider, any shooter.”

A Gotham demonstration video on Palantir’s website briefly shows a battlefield in Ukraine, along with icons that may represent military units and an AI-detected target. It is unclear whether the footage reflects Gotham’s use by Ukraine’s armed forces or is a visualization based on collected data.

Fedorov has not directly said that Ukraine’s defense forces use Gotham, but he has noted that Palantir systems are used to analyze and plan long-range strikes deep inside Russia. Since March 2026, Ukrainian forces have regularly reported (1, 2, 3) hitting strategic Russian targets located 1,500 to 2,000 kilometers from the front line.

In May 2026, a CNN report on Ukrainian forces’ use of unmanned ground and air systems revealed the existence of an intelligence and analytical software platform called PRISMA. The name of the platform is not spoken in the video, but in several shots, maps with tactical overlays can be seen in the background with PRISMA written in the upper left-hand corner.

Ukraine’s military intelligence agency (HUR) uses PRISMA for long-range drone operations. The platform brings together real-time maps, flight paths, and AI-processed data. It is capable of coordinating large numbers of drones.

The system analyzes data from previous missions — interception points, radar coverage zones, air defense activity and routes used by previous waves of drones — and uses that information to identify vulnerabilities in Russia’s air defense system and calculate alternative routes. PRISMA can process thousands of routes at once.

The CNN report also described a distributed command network, meaning that the loss of one node does not derail an entire operation. The pro-war Russian Telegram channel Arkhangel Spetsnaza (lit. “Archangel of the Special Forces”) has suggested that PRISMA is a layer built on top of Gotham, with AI algorithms adapted for the specific demands of mass raids involving dozens of drones.

For work with satellite data, Palantir developed MetaConstellation, a system that makes it possible to combine data from different satellites and quickly obtain images of a required area. With MetaConstellation, the military can quickly obtain up-to-date information about equipment movements, fortification construction, or changes on the battlefield. The Ukrainian military uses the system to integrate satellite data for reconnaissance and targeting.

Foundry is another major Palantir platform. It serves a similar function to Gotham but is designed for civilian use, acting as an AI layer that connects fragmented systems, data, processes, models, and decisions.

The platform’s users include government agencies, major corporations, and banks. In Ukraine, the Economy Ministry uses Foundry to analyze humanitarian, economic, and geographic factors in order to prioritize areas for demining efforts, while the Education Ministry uses the platform to collect and analyze data on the safety of educational institutions across Ukraine. Foundry helps create safe spaces for students, design networks of shelters, and ensure schools keep operating during wartime.

Large language models are also being used in military operations. They can process vast amounts of information, prioritize data from multiple sources, and turn a chaotic flow of intelligence into clear, high-level summaries for decision-makers.

A key part of Palantir’s ecosystem is its Artificial Intelligence Platform (AIP), which integrates large language models and other AI tools directly into military, government, and business workflows. The platform functions as a generative AI layer broadly similar to that of a chatbot: an operator asks a question, then the platform processes the request and proposes several possible courses of action, along with a forecast of the likely outcome. Palantir says AIP helps analyze large datasets, automate some analytical work, and speed up decision-making.

Another project, Brave1 Dataroom, is a secure platform created with Palantir that gives Ukrainian developers access to real battlefield datasets in order to better train AI models. Fedorov said after meeting Karp on May 12 that more than 100 companies were using the platform to train upwards of 80 models to detect and intercept aerial targets in difficult conditions.

Palantir helps collect, process, and analyze large amounts of data, but even the most accurate information must be put towards concrete action if it is to have a meaningful effect on the battlefield. Unmanned systems using artificial intelligence do exactly this.

Ukraine and Russia have entered a race to develop and deploy automated systems, from unmanned aerial vehicles to ground and naval platforms. Much of those efforts center on creating systems that can sustain operations under conditions of intense electronic warfare. Schmidt, the former Google CEO, believes the future of war will be defined by autonomous systems and the mass use of drones.

Google’s former CEO believes the future of war will be defined by autonomous systems and the mass use of drones

Schmidt led Google from 2001 to 2011, turning the company into one of the world’s leading technology firms. He later held leadership and advisory roles in the fields of defense technology and artificial intelligence, chairing the Defense Innovation Board at the U.S. Department of Defense and the National Security Commission on Artificial Intelligence.

After the start of Russia’s full-scale war, Schmidt repeatedly visited Ukraine, met with representatives of the country’s technology sector and defense industry, and said the war had become one of the main catalysts for the development of autonomous systems.

Schmidt’s Perennial Autonomy specializes in next-generation unmanned systems that combine robotics, machine vision, and artificial intelligence. Its goal is to gradually reduce drones’ dependence on operators by increasing their level of autonomy.

Unlike many traditional defense companies, the focus of Perennial Autonomy is not on creating individual weapons systems but on combining software and artificial intelligence with the mass production of drones. Schmidt argues that large networks of cheap autonomous platforms capable of quickly adapting to battlefield changes through AI will play an increasingly important role in war going forward.

Among Perennial Autonomy’s best-known systems are the Merops interceptor drone, which has proved effective against Shahed-type drones, and the medium-range Hornet kamikaze drone, which became known thanks to Ukraine’s strike campaign against the “land corridor” connecting mainland Russia to annexed Crimea. Both systems use elements of AI to strike targets.

One of Perennial Autonomy’s key technologies is autonomous drone control. Traditional FPV drones — first-person-view drones controlled through a live video feed — are entirely dependent on an operator steering the flight in real time. Autonomous UAVs, by contrast, use software algorithms that allow the drone to carry out some tasks independently, eliminating the need for constant human intervention.

Such drones can automatically follow a set route, hold a course, track a target, or complete a mission even if the link to the operator is lost. Autonomy is especially important in modern war, where electronic warfare, or EW, often disrupts drone control channels.

Schmidt advocates using large numbers of cheap drones instead of a limited number of expensive platforms. The idea is that the loss of individual drones is not critical provided that the system can quickly replace them.

That approach has proved effective in Ukraine. The mass use of FPV drones has shown that even relatively inexpensive systems can create serious problems for much more costly enemy equipment. The Hornet drone, for example, costs $10,000 to $11,000, while the targets it strikes are often worth orders of magnitude more.

Another important area is machine vision technology, one of the basic technologies involved in the development of autonomous strike drones and interceptor drones. In the case of the Hornet, this takes the form of an electro-optical guidance system. In flight, its AI processes the camera image, comparing reference images of equipment with whatever the lens is showing at that moment.

The neural network can focus on details: size, the black license plates used by the Russian military, the tracks left behind by vehicles, V and Z markings, and other small features. Based on those details, the drone determines the target’s priority and offers options to the operator, who then only has to confirm the suggested choice.

The neural network can focus on details including size, the black license plates used by the military, tracks, and V and Z markings

Autonomous systems also use AI algorithms for navigation and orientation. Such technologies allow a drone to analyze its surroundings, avoid obstacles, adjust its route, and adapt to changes in the environment — all while carrying out a mission.

Combining AI with data from cameras, sensors, and navigation systems creates the basis for more independent operation of unmanned platforms. A key feature of the Hornet UAV is its combination of satellite navigation and inertial-optical navigation. If the satellite signal is lost, the aircraft can still determine its position using images of the terrain.

A similar navigation principle was used by NASA’s Ingenuity, the first Mars helicopter. Because Mars has no satellite navigation system, the aircraft oriented itself using images of the surface below. Drones using a similar navigation principle have been nicknamed “Martians.”

An important technological challenge involves ensuring drones can operate under conditions of active electronic warfare, which suppresses control signals, data transmission, and satellite navigation. Cameras and lenses allow Hornet drones to lock onto targets from an average distance of 300 to 500 meters, allowing the UAV to dive autonomously onto a target from an altitude of 200 to 300 meters or to strike a target from roughly the same distance while flying at low altitude.

In that mode, most vehicle-mounted EW systems become ineffective. If a Hornet loses contact with its base, it does not abort the operation; instead, the drone selects a target and attacks it on its own, making it nearly insensitive to electronic warfare systems.

One of the technologies Schmidt’s company is working on is anti-aircraft UAVs. Its Merops system combines a high-speed platform, computer vision, and guidance algorithms that allow it to detect, track, and destroy aerial targets. Traditional air defense was designed mainly to counter aircraft and missiles, but the mass use of drones has forced militaries to look for new approaches. Anti-aircraft drones are a cheaper alternative to using expensive surface-to-air missiles against mass drone attacks, and interceptor drones have become part of a new layered air defense system.

The initiative showed rapid growth during Fedorov’s brief tenure as Ukraine’s defense minister. As part of the Ukrainian army’s announced technological transformation, Ukraine’s Air Force managed to increase the interception rate for Shahed drones from 83% to 91% in just six months.

Notably, immediately after Fedorov resigned as defense minister, Karp invited him to take part in a joint project, and Italian Defense Minister Guido Crosetto offered him an adviser’s post. Fedorov, however, decided to remain in Ukraine to work on technology that he believes can change the nature of modern combat.

In addition to software and artificial intelligence, scaling production has become an important part of Perennial Autonomy’s approach. Modern war has shown that the effectiveness of unmanned systems depends not only on their specifications but also on manufacturers’ ability to quickly produce them in large quantities.

If Palantir helps Ukraine work with data more effectively, Perennial Autonomy is trying to solve another problem: how to quickly turn that data into a real battlefield advantage. That is why cooperation between Ukraine and Schmidt’s company is not limited to supplying individual drones — is a long-term partnership in the production, development, and deployment of new defense technologies.

Ukraine has effectively become one of the world’s main centers for the development of military drones. No laboratory or test range can reproduce conditions like those on a real front, as developers receive constant feedback from troops using autonomous control algorithms, computer vision systems, and anti-EW solutions directly in combat.

In March 2026, Fedorov said the Defense Ministry was giving its corporate partners access to real battlefield data in order to better train their AI models. The result, he said, would benefit both sides: Ukraine receives access to new technology, while Schmidt’s company can improve its systems more quickly.

Meanwhile, Ukraine is actively developing its own systems capable of competing with global counterparts. According to Alex Karp, Ukrainian developers have created one of the most advanced and adaptive battlefield targeting systems in the world. Part of it was built with support from Palantir, but it was developed largely by Ukrainian engineers and military personnel.

The Palantir CEO sees battlefield management systems as Ukraine’s main technological asset — and, importantly, as a unique export product. Ukraine’s Delta system integrates information about enemy forces from different sources and is used both to coordinate unmanned systems and support infantry operations, allowing Ukrainian troops to use a smartphone or tablet to watch what is happening on the front line in real time, and foreign military officials have noticed. In May 2026, U.S. Army Secretary DanDriscoll called the system “absolutely incredible,” noting that it “fully integrates every drone, every sensor, every strike platform into a single network.”

Delta is an entire ecosystem, with its own messenger, cloud storage, and even an internal marketplace where units exchange electronic points in order to procure drones, communications equipment, electronic warfare systems, and other necessary gear.

Development of the system began in 2016. Today, work on it continues at the Center for Innovation and Defense Technologies Development of Ukraine’s Defense Ministry. Delta consists of several connected modules: the interactive Delta Monitor map displays the positions of Ukrainian and Russian forces in real time, Vezha (lit. “tower”)collects video feeds from Ukrainian UAVs and analyzes them using AI, Target Hub distributes targets among neighboring units, and Mission Control combines data on the activities of drone crews (including drone types, launch sites, routes, and assigned tasks).

Data gathered in Delta is used not only directly on the battlefield, but also to develop new AI systems, such as the Avengers Labs platform from the Center for Innovation and Defense Technologies Development.

Unlike Brave1 Dataroom, which is focused primarily on developing AI to detect and intercept aerial targets, Avengers Labs is designed to train models on a much broader range of battlefield data. In August, Ukraine’s Defense Ministry said Ukrainian defense companies had received access to the platform, which contains around 5 million annotated frames, mostly obtained from the Delta ecosystem. They cover a wide range of targets, including tanks, artillery, air defense systems, individual infantry soldiers, and strike and reconnaissance drones. Companies can use this real combat data to train their own AI models without receiving direct access to sensitive information. Such models are expected to be used, among other things, for automatic target recognition and autonomous drone guidance.

In March 2026, Ukraine’s Defense Ministry launched the Defense AI Center A1 for battlefield analysis and technology, which then-Defense Minister Fedorov said would “help turn combat experience and front-line data into technological solutions more quickly and accelerate the introduction of innovation into the troops.” The center’s main tasks involve analyzing combat data, forecasting enemy actions, developing autonomous systems, and creating new command-and-control tools.

According to the center’s head, Danylo Tsvok, artificial intelligence is already helping Ukraine hold territory and reduce risks to soldiers as they confront a larger and better-resourced enemy. As the technology develops, AI could become the foundation of a single digital battlefield in which intelligent weapons systems act in coordination, relying on a shared platform for command and battlefield analysis.

“This could happen within three to five years,” Tsvok said. “Within that timeframe, the front line could be protected by tightly integrated hardware and software systems.”

Artificial intelligence is already helping Ukraine hold territory while reducing risks to soldiers

Fedorov believes that, given the nature of modern warfare, Ukraine must “defeat Russia in every technological cycle, and artificial intelligence is one of the key areas of this competition.”

“The world needs security, and only autonomous weapons can provide it,” the former minister said. “This is the new nuclear weapon. Countries that have it will be protected.”

Eric Schmidt believes that in future military conflicts, the front line will become a digital-age version of the “no man’s land” famous from World War I — an area where sensors and drones will make it possible to strike everything that moves.

Schmidt believes combat AI systems will become increasingly autonomous — humans will transition from controlling every shot to merely exercising broad supervision, intervening only when necessary. Two primary factors are driving that shift: the regular loss of contact with drones because of electronic warfare and the battlefield advantage held by more autonomous systems over those in which every use of force requires operator confirmation.

Schmidt believes this logic, like the chemical weapons race in World War I, could push states toward the mass adoption of autonomous combat AI systems, even if they initially would have preferred stricter limits on the use of such technologies.

Combat AI systems will become increasingly autonomous

Modern war has changed radically: the mass use of cheap drones is proving more effective than a limited number of expensive weapons systems. But the armies and defense industries of many countries, including the United States, continue preparing for past conflicts, relying on costly and complex platforms even as the experience of the war in Ukraine shows the effectiveness of different approaches. A massive revolution in military affairs is underway, but political and military leaders have not yet adapted to the new conditions, Schmidt stresses.

Whether the pace of Ukraine’s technological transformation will continue after Fedorov’s departure from the government remains an open question. According to a Conflict Intelligence Team (CIT) expert who asked to remain anonymous, many of the reforms that are already underway have enough momentum to continue. But at the same time, the expert said, many initiatives were closely tied to Fedorov’s personal role — his professional contacts, his talent for building effective teams, and his ability to secure support for projects at various levels.

The expert also pointed to other personnel changes, including the departure of Pavlo Yelizarov as deputy commander of Ukraine’s Armed Forces, which could slow some projects or lead to their cancellation. Still, the expert emphasized that it is too early to draw firm conclusions. Much will depend on the final makeup of the new leadership at Ukraine’s Defense Ministry and the priorities set by the new minister.

Military analyst Kirill Mikhailov takes a different view. He believes the technological transformation of Ukraine’s defense sector has already become irreversible and is unlikely to be stopped by personnel reshuffles. At the same time, he links the future of battlefield AI primarily to expanding commanders’ ability to analyze the situation and make decisions. As Mikhailov notes, modern AI systems can be more effective in command-support roles than some staff officers.