Artificial Intelligence (AI) in Modern Warfare: Ukraine, Gaza, Iran.

 Between 2022 and 2026, three Armed conflicts have served as the operational test bed for AI in war. Ukraine shows how cheap,expendable autonomous hardware spreads at scale. Gaza shows how AI moves into the cognitive core of military targeting. The Israel-Iran exchanges show how AI now coordinates campaigns across multiple domains at machine speed.

The war in Ukraine is the most consequential testing ground for autonomous and AI-enabled systems in modern combat. By December 2025, Defence Minister Denys Shmyhal said Ukraine’s Defence Procurement Agency had supplied 2.4 million firstperson-view drones in a year, with total armed forces uptake on track for three million units. Ukrainian production capacity is reported to have reached as many as five million drones in 2025, drawing on more than 500 domestic manufacturers via the Brave1 defence-tech accelerator launched in April 2023. Figure 5.2 shows drone strike events and resulting fatalities from 2018 to 2025; Ukraine is separated from the rest of the world to underscore its disproportionate contribution to global drone warfare. Recorded drone strike events rose more than a hundred-fold across the period, climbing from 364 in 2018 to over 42,000 in 2025. The pace of expansion accelerated sharply after Russia's invasion of Ukraine in 2022, and Ukraine's share of global activity has risen in every subsequent year.

Across the full 2018–2025 window, Ukraine accounted for almost 50 per cent of all drone strike events, a higher share than any other country. By 2025 that figure had reached 57 per cent. Russia is the next-largest single jurisdiction by event count, at 30.6 per cent. Almost all of these are Ukrainian deep-strike operations against Russian territory rather than a separate conflict. Taken together, the Russia-Ukraine war therefore accounts for roughly four out of every five drone strikes recorded worldwide since 2018. The fatalities panel tells a different story. While events are heavily concentrated in Ukraine, the lethality per strike has been much lower than in other theatres. Ukraine's share of cumulative drone strike fatalities is 16.5 per cent, well below its share of events, and Palestine is the most lethal theatre, accounting for 22 per cent of all drone strike fatalities despite accounting for only 2.6 per cent of events. Sudan, Myanmar, and Syria together add another 30 per cent of fatalities, underlining that drone use in active civil wars tends to produce higher per-event casualties. Several Ukrainian systems now use machine vision and onboardAI in the final phase of a strike. The Saker Scout, approved for armed forces use in September 2023, can identify 64 categories of Russian military equipment and run autonomous strikes after losing GPS or radio link. It works with Delta, Ukraine’s NATO-compatible situational-awareness platform, and GIS Arta, the artillery-fire-control system credited with cutting target-tofire times from roughly 20 minutes to under one minute.

 A growing fleet of unmanned ground systems, including the Lyutmini-tank, the Termit modular ground vehicle, and the devDroid family, now field AI-assisted target detection alongside conventional teleoperation. Brave1 has also incubated Griselda, an intelligence-fusion platform that combines open-source, signals, and battlefield reporting for commanders. On the Russian side, the V2U loitering munition, first publicly identified by Ukraine’s Main Intelligence Directorate in June 2025, shows how fast the threshold of meaningful human control is shifting. Analysis of intercepted V2U airframes confirmed that the drone uses an Nvidia Jetson Orin chip on a Chinese carrier board to autonomously find and select targets, with no operator link once airborne. Russia has likely fielded afully autonomous combat drone and continues to iterate despite civilian casualties. In one May 2025 incident, seven V2U units reportedly broke off a planned mission, autonomously formed a holding pattern, and coordinated attacks on a column of vehicles and civilians. By mid-2025, Ukrainian intelligence estimated Russia was launching 30 to 50 V2U sorties a day. There have also been significant advances in drone usage by air force and navy units. On 31 December 2024, a Ukrainian Magura V5 unmanned surface vessel operated by HUR Group 13 became the first sea drone to shoot down a manned helicopter, downing a Russian Mi-8 near Cape Tarkhankut in occupied Crimea with R-73 air-to-air missiles. In May 2025, a successorMagura V7 destroyed two Su-30 jets near Novorossiysk, the first time an unmanned vessel downed a fixed-wing aircraft. For most countries to date, drones are an important military innovation, but the revolution they represent is potential, not yet realised. What gives Ukraine its global significance is the trajectory toward fully autonomous targeting, swarm coordination, and machine-speed engagement. Major militaries have responded to the success of drone warfare in Ukraine and are institutionalising the approach. The US Department of Defense’s Replicator initiative, announced by Deputy Defense Secretary Kathleen Hicks in August 2023, set out to field thousands of low-cost and expendable autonomous systems across multiple domains within 18 to 24 months.16 Replicator 2, announced in late 2024, focused on counter-smallUAS defence of US installations.17 In August 2025, Defense Secretary Pete Hegseth consolidated counter-drone activities under Joint Interagency Task Force 401. At Camp Atterbury, Indiana, that same month, Epirus’s Leonidas high-power microwave system defeated 61 of 61 drones across five scenarios, including a 49-drone swarm taken down by a single pulse. Indo-Pacific Command’s Admiral Samuel Paparo summed up the strategic logic in June 2024.20

If Ukraine shows AI hardware spreading at scale, Gaza shows AImoving into the cognitive core of military targeting. Two systems disclosed in +972 Magazine and Local Call have anchored the global debate. The first, Habsora (‘The Gospel’), was described in November 2023 by an IDF spokesperson as a system that uses automatic tools to produce targets at a fast pace. Former IDF Chief of Staff Aviv Kochavi told Ynet in June 2023 that once the system was activated, it generated 100 new targets a day, against a previous baseline of around 50 a year. The second system, Lavender, was disclosed in April 2024. According to one investigation, during the first weeks of the war the army almost entirely relied on Lavender, which flagged as many as 37,000 Palestinians as suspected militants. Internal testing reportedly produced a 90 per cent accuracy rate, treated as sufficient for sweeping operational use, with the other 10 per cent known to be misidentifications. Human operators often served as a rubber stamp, spending around 20 seconds per target before authorising a bombing, just to confirm the target was male. A companion system, Where’s Daddy?, tracked individuals on the Lavender list so strikes could happen when they returned home at night. For junior operatives, commanders permitted up to 15 to 20 civilian deaths per strike.  A joint investigation published in August 2025, drawing on a leaked Israeli military intelligence database, found that as of May 2025 it listed 8,900 Hamas and Palestinian Islamic Jihad fighters as dead or probably dead. Set against around 53,000 Palestinian deaths recorded by Gaza health authorities at the time, the named-fighter share stood at roughly 17 per cent. Independent modelling puts the total much higher: a November 2025 study estimated 78,318 violent deaths between October 2023 and December 2024, with later updates suggesting cumulative conflict-related deaths in Gaza had likely passed 100,000 by October 2025 and that life expectancy had been cut by more than 30 years over two consecutive years. Similar targeting systems are being implemented by other advanced militaries. In the US, Project Maven, originally a 2017 Pentagon computer-vision pilot, has become a fully institutionalised combat platform. In May 2024, the Pentagon awarded Palantir a five-year, US$480 million contract to expand the Maven Smart System. By May 2025, the contract ceiling had risen by US$795 million to roughly US$1.3 billion, a 165 per cent expansion in 12 months. US Central Command director of operations Schuyler Moore told Bloomberg that after the October 2023 attacks, Maven had become exceptionally critical to function, with targeting tempo rising sharply

The third conflict where AI has played a crucial role in the war in Iran. The June 2025 Twelve-Day War, in which Israeli and US strikes targeted Iranian nuclear facilities, was followed in February 2026 by the larger Operation Epic Fury, a coordinated multi-domain US and Israeli campaign against Iranian leadership, missile, and nuclear infrastructure. According to US and Israeli statements, 200 Israeli aircraft struck around 500 targets and US forces conducted 900 strikes in the first 12 hours; Iran responded across nine countries with several hundred missiles and drones. A Hudson Institute analysis found that Iranian missile salvos decreased 86 per cent within two weeks, attributed to the systematic destruction of mobile missile launchers identified through persistent surveillance and machine-vision target classification. Operation Epic Fury can thus be seen as an inflection point in how AI is being integrated into operational planning, kill-chaincompression, and contested airspace. The wider implication is that a campaign run at machine speed across multiple theatres, with humans validating rather than originating most targeting decisions, sets a new operational baseline. Military capabilities are evolving faster than the legal and governance frameworks meant to regulate them, and the role of human judgement within these systems is becoming less certain.

The three Armed Conflicts theatres show three distinct ways in which AI is being integrated into armed conflict. Ukraine shows cheap autonomous hardware at scale, with a tight feedback loop between civilian manufacturing and battlefield iteration. Gaza shows AI-driven cognitive targeting and the compression of human review to a window measured in seconds. The IsraelIran exchanges show AI cueing across multiple domains at machine speed.

The combined impact of the advances in AI in warfare can be seen in Figure 5.3, which shows the evolution of target-to-fire times of key military targeting systems over the past 30 years.Advances in AI have led to significant ‘kill-chain compression’: targeting decisions that once took hours or days can now occur in a matter of minutes, or even seconds.