‘Attacks will be fully autonomous’: Russia, Ukraine race towards AI warfare

Automated AI drones are just one component of an ultimate chain of destruction both sides are building.

Drones with some autonomous capabilities are already being deployed at scale by Russia and Ukraine [File: Daryna Naumova/EPA]

Tetiana Bubynets collapsed on the asphalt after a Russian drone exploded at a gas station in the southern Ukrainian city of Zaporizhzhia.

Despite the violent shockwave and blast of drone shrapnel produced by the attack on July 6, the 18-year-old accounting student had no visible wounds except for a nosebleed.

list 1 of 3 US envoys meet Putin: What’s behind latest diplomacy on Russia-Ukraine war?

list 2 of 3 US envoys hold ‘substantive’ talks with Zelenskyy in first visit to Kyiv

list 3 of 3 Witkoff says peace talks have been ‘very meaningful’ in Ukraine

She was rushed to hospital and pronounced dead shortly after, alongside two other civilians, aged 41 and 48.

One of her lecturers, Tetiana Melikhova of Zaporizhzhia National University, said Bubynets was an only child and that her mother “lost the meaning of life” after she died.

“She couldn’t speak. At the funeral, she just lay on the coffin; she couldn’t let her child go,” Melikhova told Al Jazeera.

The strike that killed Bubynets was the first recorded case of Russia’s use of a fully autonomous attack drone powered by artificial intelligence (AI), according to a New York Times analysis .

While drone strikes have hammered Ukraine since Russia launched its full-scale invasion in February 2022, Ukrainian officials who examined the wreckage found the unmanned vehicle to be unlike others.

In place of an antenna, it featured an AI-powered minicomputer sold commercially by US technology company Nvidia, which rendered it capable of directing itself to the type of target it was programmed to spot – likely propane tanks in this case – lock on to it and proceed to strike.

AI-powered drones are the tip of the iceberg as Russia and Ukraine race towards boosting tech infrastructure aimed at coordinating reconnaissance and attack, according to analysts.

“It’s not about a drone having AI in it, and then seeing something and striking,” Marina Miron, researcher at the defence studies department at King’s College London, said. “Rather, it’s the drone seeing something, communicating to a software complex which knows how to map the GPS coordinates on a live battlefield map, and notifying the closest artillery unit.”

Referred to as reconnaissance-strike architecture, these unified battlefield management systems act like a central command. They track the location of friendly forces and identify enemy positions.

AI acts as the accelerator, instantly processing massive amounts of data collected through sensors in drones or any other military products, including interceptor missiles or artillery units.

The tactical advantage on the battlefield is given by the system’s ability to compress the kill chain – meaning coordinating and distributing the intelligence to military assets on the ground to find, track and destroy targets in a matter of minutes.

In such an ecosystem, AI-powered automated drones are just pieces of a broader puzzle.

Ukraine is known to be developing its system, which it named DELTA. Following operational trials in late 2025, Russia has been rolling out its version, the Svod Tactical Situational Awareness Complex, to front-line units.

Both sides claim that once the battlefield data is processed, a decision support system (DSS) presents a military commander with options to choose from, keeping a human in the decision-making loop, said Miron.

“In how far that is true is, of course, impossible to find out,” she added.

Acknowledging full AI autonomy would provoke ethical and legal questions.

Even if humans were to retain final decision-making authority, research has shown AI systems subtly shape choices, predetermining outcomes.

“Under time and pressure, commanders often do as AI suggests,” said Miron, whose research focuses on military AI and remotely piloted and autonomous systems.

Towards full automation

Coordinated AI-powered military systems are the next frontier of modern warfare, said Serhiy Beskrestnov, a Ukrainian Ministry of Defence adviser on loitering munitions, known by his callsign Flash.

“That’s what it will all come to,” he posted on his Telegram channel on July 1. “Navigation, search for a target, and attacks will be fully autonomous.”

Drones with some autonomous capabilities are already being deployed at scale.

Heorhii Volkov, commander of Ukraine’s Yasni Ochi drone strike unit, said this is currently mostly used in the final phase of long-distance strikes, where first-person view (FPV) drones are likely to lose radio connection due to electronic warfare or the curvature of the Earth.

The use of AI in so-called last-mile operations solves this battlefield hitch. If the signal is lost, an integrated AI-enabled onboard module takes full control of the unmanned vehicle, autonomously locks on the target and carries the strike to completion.

Yaroslav Azhnyuk, founder of Ukrainian technology company The Fourth Law, which develops AI-powered autonomy software for unmanned vehicles, said Russia and Ukraine were working to deploy a wider range of autonomous systems.

“This is not simply a race to demonstrate a fully autonomous drone. It is a race to make autonomy reliable, affordable and deployable at scale,” Azhnyuk told Al Jazeera. “Electronic warfare, increasingly long engagement distances and the growing number of drones are making human-controlled systems less sufficient. Autonomy solves exactly these constraints.”

Azhnyuk believes autonomy will become a game-changer over the next six to 12 months.

“The decisive advantage will belong to the side that can deploy reliable autonomy across tens or hundreds of thousands of systems, not just demonstrate it in prototypes,” he said.

Kateryna Bondar, senior fellow at the Center for Strategic and International Studies (CSIS), noted the dangers of AI’s reliance on algorithms.

“The system doesn’t understand what it’s doing; it’s just following the algorithm and the rules which people created,” said Bondar, who works at the CSIS’s Wadhwani AI Center. “AI has to be trained to recognise targets and engage these targets,” but writing code that distinguishes between a bus and a tank is easier than between a Russian and a Ukrainian soldier.

Additionally, cost-effective FPV drones have low-quality cameras and their resolution can be lowered further by adverse atmospheric conditions.

Military units using autonomous systems therefore “delegate the decision to a very imperfect technology that cannot distinguish between a civilian and a military person, combatant or non-combatant”, Bondar said.

“This is why there is a conversation around who makes a decision and who has responsibility for mistakes and errors.”

自动化人工智能无人机只是双方正在构建的终极毁灭链条中的一个组成部分。

俄罗斯和乌克兰已经开始大规模部署具备一定自主能力的无人机。[图片：Daryna Naumova/EPA]

在乌克兰南部城市扎波罗热的一个加油站，一架俄罗斯无人机爆炸后，塔蒂亚娜·布比涅茨倒在了沥青路面上。

尽管7月6日的袭击产生了强烈的冲击波和无人机碎片爆炸，但这位18岁的会计系学生除了流鼻血外，没有其他可见的伤口。

三位美国特使会见普京名单 1：俄乌战争最新外交举措背后有何玄机？

美国特使三人中的第二人首次访问基辅，与泽连斯基举行“实质性”会谈。

维特科夫表示，乌克兰的和平谈判“意义重大”（共3条记录，此为第3条）。

她被紧急送往医院，不久后被宣布死亡，一同遇难的还有另外两名平民，年龄分别为 41 岁和 48 岁。

她的一位讲师，扎波罗热国立大学的塔蒂亚娜·梅利霍娃说，布比涅茨是独生女，她的母亲在她去世后“失去了生活的意义”。

“她无法说话。葬礼上，她只是躺在棺材上；她舍不得自己的孩子离开，”梅利霍娃告诉半岛电视台。

据《纽约时报》分析，导致布比涅茨丧生的那次袭击，是俄罗斯首次使用由人工智能 (AI) 驱动的完全自主攻击无人机的记录案例。

自俄罗斯于 2022 年 2 月发动全面入侵以来，无人机袭击一直重创乌克兰。但检查残骸的乌克兰官员发现，这架无人驾驶飞行器与其他飞行器不同。

它没有天线，而是配备了美国科技公司英伟达（Nvidia）商用的人工智能微型计算机，使其能够自主导航至预先编程的目标类型（在本例中可能是丙烷罐），锁定目标并进行攻击。

分析人士称，人工智能无人机只是冰山一角，俄罗斯和乌克兰正在竞相加强旨在协调侦察和攻击的技术基础设施建设。

伦敦国王学院国防研究系的研究员玛丽娜·米隆表示：“关键不在于无人机搭载人工智能，然后发现目标并进行攻击。而在于无人机发现目标后，与能够将GPS坐标映射到实时战场地图上的软件系统进行通信，并通知最近的炮兵部队。”

这些被称为侦察打击架构的统一战场管理系统，如同一个中央指挥中心。它们能够追踪友军位置并识别敌军阵地。

人工智能充当加速器，即时处理通过无人机或其他军事产品（包括拦截导弹或火炮部队）中的传感器收集的大量数据。

该系统在战场上具有战术优势，因为它能够压缩杀伤链——这意味着协调和分发情报给地面军事力量，以便在几分钟内找到、跟踪和摧毁目标。

在这样的生态系统中，人工智能驱动的自动化无人机只是更大拼图中的一块。

乌克兰正在研发一套名为“DELTA”的系统。俄罗斯在2025年底完成作战试验后，已开始向前线部队部署其版本——“Svod”战术态势感知系统。

米隆表示，双方都声称，一旦战场数据得到处理，决策支持系统 (DSS) 就会向军事指挥官提供多种选择，从而使人参与到决策过程中。

“当然，我们不可能知道这在多大程度上是真实的，”她补充道。

承认人工智能完全自主性会引发伦理和法律问题。

即使人类保留最终决策权，研究表明人工智能系统也会微妙地影响选择，预先决定结果。

“在时间和压力下，指挥官们往往会按照人工智能的建议行事，”米隆说道，他的研究重点是军事人工智能以及远程驾驶和自主系统。

迈向全面自动化

乌克兰国防部巡飞弹顾问谢尔盖·贝斯克列斯特诺夫（代号“闪电”）表示，由人工智能驱动的协同军事系统是现代战争的下一个前沿领域。

“最终都会变成这样，”他在7月1日于自己的Telegram频道上发帖说。“导航、搜索目标和攻击都将完全自主化。”

具备一定自主能力的无人机已经开始大规模部署。

乌克兰“亚斯尼·奥奇”无人机打击部队指挥官赫奥尔希·沃尔科夫表示，目前这种技术主要用于远程打击的最后阶段，因为第一人称视角（FPV）无人机可能会因电子战或地球曲率而失去无线电连接。

在所谓的“最后一公里”作战中运用人工智能可以解决战场上的这一难题。如果信号丢失，集成的人工智能机载模块将完全控制无人车辆，自主锁定目标并完成打击任务。

乌克兰科技公司“第四法则”（The Fourth Law）的创始人雅罗斯拉夫·阿日纽克（Yaroslav Azhnyuk）表示，俄罗斯和乌克兰正在努力部署更广泛的自主系统。“第四法则”是一家为无人驾驶车辆开发人工智能驱动的自主软件的公司。

阿兹纽克告诉半岛电视台：“这不仅仅是一场展示全自主无人机的竞赛，更是一场让自主技术可靠、经济且可大规模部署的竞赛。电子战、日益远距离的交战以及无人机数量的增长，都使得人工控制系统显得力不从心。而自主技术恰恰可以解决这些难题。”

阿兹纽克认为，在未来六到十二个月内，自主性将成为改变游戏规则的关键因素。

他说：“决定性的优势将属于能够在成千上万个系统中部署可靠自主性的一方，而不仅仅是在原型机上进行演示的一方。”

战略与国际研究中心 (CSIS) 高级研究员 Kateryna Bondar 指出，人工智能依赖算法存在危险。

“这套系统并不理解自己在做什么；它只是遵循人为设定的算法和规则，”邦达尔说道，他是美国战略与国际研究中心（CSIS）瓦德瓦尼人工智能中心的研究人员。“人工智能需要经过训练才能识别目标并攻击这些目标，”但编写能够区分公共汽车和坦克的代码比区分俄罗斯士兵和乌克兰士兵要容易得多。

此外，性价比高的FPV无人机配备的摄像头质量较低，恶劣的大气条件还会进一步降低其分辨率。

邦达尔说，因此，使用自主系统的军事单位“将决定权委托给一种非常不完善的技术，这种技术无法区分平民和军人、战斗人员和非战斗人员”。

“这就是为什么大家会讨论谁来做决定，谁来承担错误和失误的责任。”