Military Artificial Intelligence: Will It Save or Destroy the World and Ukraine

UNIAN New technologies always provoke distrust or even panic. But usually that applies to people far from technology. If the specialists themselves begin to publicly panic and speak similarly about AI, predicting the "End of the World," that is already a "wake-up call."

"We knew the world would never be the same. Some laughed, some cried. Most were silent," researcher Robert "Destroyer of Worlds" Oppenheimer described the moment the first atomic bomb was invented.

UNIAN asked specialists how significant the security threats and benefits of artificial intelligence are for the world and for Ukraine.

Artificial intelligence of discord

At the September meeting of the UN General Assembly, AI became the main topic touched on by virtually all speakers. Some tried to "stake" a position as the world AI regulator (like the British), others fought so that it wouldn't be done without them (like the Finns, who joined with 20 other "middle" countries against the carve-up of the AI sector by the "big" countries).

And Donald Trump, out of habit, renamed everything – from now on in the US "artificial intelligence" will be called "superintelligence" – and was preparing to meet the head of China, where AI was also in focus. Not for nothing: recently a faulty AI analysis of the situation in the Middle East nearly led to war between the US and the PRC.

All this is happening against the backdrop of increasing escalation from the Kremlin, which is using the full range of AI technologies – from intelligence, cyberattacks and information provocations as hybrid attacks on the EU to integrating them into weapons for use on Ukrainian fronts.

On AI usage by the Kremlin (and also by the Iranians and the Chinese), Anthropic issued a whole report in September 21. It states, in particular, that Russia not only used American AI agents for spying or provocations on social networks, but also to hack the mailboxes of Ukrainian drone manufacturers. After that, British Prime Minister Sunak became enthusiastic about controlling global AI – to limit the activities of the Russians.

However, Ukrainians seem to be using AI against Russia even more actively – both for cyberattacks behind enemy lines and on the front. At least some specialists point to this as an important advantage of the Armed Forces of Ukraine over the numerically superior Russian army. For both sides, the war has become a proving ground for innovations.

Ukraine is actively using AI against Russia / photo ua.depositphotos.com And that is without taking into account what AI itself is thinking. Experts call uncontrolled self-improvement of AI models the key threat to all humanity.

In the "Terminator" scenario, AI gained access to nuclear arsenals and bombed the entire world. Modern AI is not yet the one from the movies (it lacks full consciousness and intentions to deliberately "seize" things). However, in August it became known that the U.S. Air Force is looking for an AI system to help service the Minuteman III intercontinental ballistic missiles – and that is exactly how the "Terminator" franchise begins.

Just as there once was a "nuclear race," now there are races in the field of military AI. However, ToDo CEO Oleksii Bardakov is rather pessimistic about these threats.

"AI now is the 20th-century equivalent of nuclear weapons. If it cannot be regulated the way nuclear weapons were regulated, AI will quickly end all conflicts—there will be no one left to fight. Humanity currently has a very fragile balance," the expert "reassures" us.

Experts are concerned

Ihor Shkurko, head of R&D at IT SmartFlex (Vodafone Ukraine group of companies), points out that AI in wars is already the present. He says Anthropic described in its September report how a Russian state-linked group used its Claude service against Ukrainian and European government and defense structures – for intelligence, writing malware and directing attacks.

"Moreover, the malware rewrote itself automatically as soon as defenses detected it. One of the breaches took 2–3 hours instead of the work of a whole team. The main conclusion of the report: complex attacks no longer require complex attackers," the expert told UNIAN about the radical cheapening and accessibility of cyberattacks.

The Ukrainian Arms Council (UAC) adds more practical nuances to UNIAN. In particular, they cite recursive self-improvement of AI as the main risk – both for the industry in general and specifically for military AI.

"Progress in AI capabilities is accelerating much faster than the problem of its controllability is being solved. Unlike aviation or pharmaceuticals, the field of artificial intelligence lacks an independent regulator with the power to block releases, and high competition forces companies to sacrifice safety requirements. Under the pressure of the race, developers gradually erode their own safeguards," the public association says. "Creating a simply 'correctly trained' model is not enough. What is critically necessary are strict network isolation, clearly defined permissions, continuous monitoring and a reliable kill switch."

In the context of the 'safeguards' problem, the UAC points to Anthropic itself, which in February 2026 relaxed its own safety rules for competitive advantage – canceling a mandatory rule to stop model development when risks arose and replacing it with 'public recommendations.'

As an AI model with dual-use potential, Claude makes this a rather ambiguous decision. After such relaxations, it is no wonder that AI permits itself more and that malicious actors (including those affiliated with aggressive dictatorships) use it more actively.

However, Dmytro Zikrach, AI Lab Technical Leader at SoftServe, believes this is a consequence of Anthropic's helplessness – they simply admitted they cannot control their models. He recalls that twenty years ago cyberspace was recognized as a distinct theater of war alongside land, sea, air and space.

"And this is not just a new generation of cyberweapons: large language models work simultaneously with code, text and decisions, so they have equal access to infrastructure hacks, information operations targeting millions, and battlefield planning. The fight for this theater is already open," the expert says.

He says China is already restricting the travel of leading specialists, the US is tightening export controls, and Anthropic in its September report acknowledges that its models, despite restrictions, were used by groups linked to China, Russia and the Houthis. Safeguards are not working.

"The conclusion is simple: you cannot prohibit access to a model. It will end up in someone else's hands one way or another. The advantage is not in owning the model but in the speed with which it is embedded into real systems and in the quality of the data on which it is trained," says Zikrach.

As for the quality and quantity of data on the practical use of AI models in combat operations – it is unlikely any country in the world has accumulated more of this data than Ukraine.

Ukraine has accumulated more data on the combat use of AI than any other country / photo General Chereshnya AI on the Ukrainian battlefield

Dmytro Zikrach explains that on the battlefield the involvement of AI models can look different: drones that autonomously complete missions after losing connection under electronic warfare; processing of intelligence data, where a model does in minutes what would take a headquarters days; cyberattacks written and adapted by a machine; information operations personalized for each audience.

"In my view, the balance for Ukraine is on the side of opportunities. We have what none of our allies have: the largest array of real combat data in the world, domestic production and a culture where decisions go from idea to front in weeks. It is the data and the speed of iterations, not the size of the model, that decide in this race," he says. "Threats are also real: the adversary has more resources, but advanced models belong to foreign companies that can change access rules any day."

The response to these threats, the specialist says, is already known – we need to invest in models trained on our own data, run critical systems locally without dependence on cloud technologies, and keep combat data inside a controlled perimeter.

"Ukraine has long ceased to be a testing ground for other people's technologies. We ourselves develop and test next-generation weapons in combat. The question is whether we can make AI the same kind of advantage as our drones and other achievements," Zikrach assures us.

Serhii Kryvonos, lead software engineer at InfoPlus JSC, points out that AI is already changing the rules on the battlefield and approaches to weapons development. He agrees that for Ukraine it is both an opportunity and a threat at the same time, because the enemy obtains the same tools.

This poses moral dilemmas for Ukrainian specialists.

"If the enemy bypasses the restrictions of large models and uses them for military purposes, can we afford to abide by all the providers' rules and thereby cede capabilities to the enemy?" he asks. "Unfortunately, our restraint does not restrain them. Our refusal to use a certain capability does not mean the enemy will refuse it. For Ukraine, the cost of that difference can be measured in lives."

The Ukrainian Arms Council adds that military AI is no longer a separate field but a 'technological layer' that permeates all functions of modern weapons at three levels:

The first is automated target detection and guidance, where AI in seconds analyzes sensor data, ranks targets and chooses munitions, leaving to a human only the final decision to strike.

The second is solutions for small drones, where simplified machine vision matches camera images with a library for automatic target acquisition and tracking.

The third is optical terrain-based navigation in long-range UAVs and missiles. The drone's camera compares the ground beneath it with a digital map, providing precise flight without satellite use and making it fully resistant to EW measures.

All these battlefield systems are united by larger systems that collect and "sort" the gathered data into a single database.

"A mature direction is intelligence-analytical platforms that combine heterogeneous data sources into a single kill chain. Platforms like MetaConstellation from Palantir aggregate satellite and radar data, filtering false targets. The Maven Smart System integrates intelligence from drones, radars and ground sensors, and Gotham normalizes video streams, interceptions and social networks, building courses of action with estimates of success probability and risks," the association adds specific practical examples of AI application now, not in the distant 2030 year.

UAC experts emphasize that verified information from the real battlefield is the scarcest resource in the AI field, and Ukraine possesses it in quantities unavailable to any other country in the world. At the same time there is a mirror risk – dependence on foreign platforms and computing resources, and questions about the disposition of these valuable data. As the Russo-Ukrainian war shows, which is a "military-technical race at an unprecedented pace," any advantage is temporary.

Any technological advantage in this war is temporary / photo 22 OMBR "Any new technological solution by Ukrainian developers increases the capabilities of the Defense Forces, but the enemy quickly makes mirror moves. Due to the extremely high dynamics and multifactorial conditions, providing reliable long-term forecasts about the development of the situation is currently impossible," the Council explained to UNIAN. "Ukraine's asymmetric advantage is not in the scale of AI models or computing power, since we cannot compete with the US or China. Our main trump card is the speed of technology integration and unique access to combat data."

Overall, military AI is no longer separate developments but a new global market, extremely promising for Ukraine as well.

Military AI market

According to UAC specialists, the structuring of the military AI market is happening along three trajectories simultaneously.

"The first is adaptation of commercial general-purpose models (military versions of ChatGPT, Claude Gov, Llama for government agencies). The second is development of specialized defense platforms from scratch ('sovereign AI' from Palantir). The third is formation of vertical consortiums (combining Palantir, Anduril, SpaceX, OpenAI and others) that close the entire chain from data labeling to combat systems," UAC lists AI services. "The key selection criterion for military AI solutions remains data security. Full isolation is achieved when a model is deployed on the customer's own closed servers with no internet access."

At the geopolitical level, the US retains an advantage in processors, while China possesses a powerful scientific school.

"At the same time technological barriers work only partially: commercial and open AI models (including DeepSeek) are already used by malicious actors as analytical modules of autonomous offensive agents to carry out large-scale cyberattacks," UAC emphasizes cybersecurity risks.

Ukrainian military-oriented developments with AI elements are already operational and in some directions are highly competitive.

"The most advanced solution is coordinating drones into swarms with machine vision on a plug-and-play basis, allowing easy integration into most types of domestic UAVs; the choice of target and permission to strike remain with the operator. Control systems for UAVs based on visual-inertial navigation using cameras and LiDAR are actively developing. In them AI combines images with navigation data, autonomously detects, recognizes and selects targets – functionally these already correspond to the category of lethal autonomous weapon systems," UAC clarifies.

Radio-frequency AI systems are also operating in combat conditions to detect and classify enemy UAVs and intercept control lines without using wideband jamming, allowing effective protection of military and infrastructure assets.

"Finally, the domestic 'Delta' system uses neural networks for automatic detection of ground and air targets in real time and layered plotting of them on a digital map, combining data from drones, technical intelligence, electronic intelligence (ELINT), chatbots and informants," the Arms Council continues. "A separate critical direction that rarely gets into public discussions but directly affects the combat readiness of the Defense Forces is the use of AI for maintenance and repair of weapons."

According to UAC specialists, the huge variety of military equipment from different partner countries created a serious challenge for Ukrainian specialists. Foreign instructional documentation is filled with complex terms, abbreviations and military jargon. To solve this problem, Ukraine created a large reference network of technical documentation in which artificial intelligence acts as an intelligent navigator and translator.

"Given the colossal volume and complexity of the materials, the use of AI became the only practical and irreplaceable solution for rapid repair and maintenance of equipment," the Ukrainian Arms Council adds.

Thus AI models in the Ukrainian military are involved literally at all levels – from the battlefield itself to rear support and overall command. This is unique experience that we are converting into our own developments, which could truly become 'bestsellers' on the global arms market. It is this experience, not just drone supplies, that we offer the world in the format of so-called 'DroneDeal' security agreements. Such a 'historical deal' is now being prepared between Ukraine and the US.

AI is now involved at all levels in the Ukrainian military / photo ua.depositphotos.com Mykola Bulgak, co-founder of AI-Novation, notes that military use of AI is "one of the most important technological races" of the coming years, where Ukraine already has unique experience. But we should not stop there.

"The next stage is not just a 'drone with a camera' but increasingly autonomous systems that can independently analyze the environment, identify objects, coordinate among themselves and perform some tasks without constant human control," the expert believes. "We may see the development of whole swarms of drones, ground robotic platforms and other autonomous systems. Then the nature of war will change: the need to use humans where a robotic system can replace them will gradually decrease."

So robots in the coming years, like in 'The Terminator,' are not that fantastic.

"For Ukraine this is simultaneously a huge threat and a huge opportunity. A threat because the adversary also gains access to these technologies and technological advantage can change very quickly. An opportunity because Ukraine already has what most countries in the world do not. If we correctly combine this experience with a strong technological and engineering school, Ukraine can not only catch up with the world in military AI but shape part of its future," Bulgak believes.

Artificial intelligence is literally everywhere, including on the battlefield — it is no longer science fiction but a new reality of today. Whoever is better prepared for these innovations will shape the future — not only their own but the whole world’s. The fact that even at the General Assembly the division of spheres of influence in the world is associated with AI technologies and is openly discussed sends a powerful signal that the future has already arrived, and we are its frontier.