Loitering Munitions and Civilian Impact: The Rise of Autonomous Warfare

Co-authored by Saankya Suresh, Ani Tutunjyan, Dune Strickland, Olivia D’Anna, Tasmia Puspita, Tyrel De Veyra

Russian drone Geranium, based on Iran’s Shahed, found in Vinnytsia region of Ukraine in March 2024. Image by the National Police of Ukraine via Wikimedia Commons.

In 2025, the US loitering munitions market size was $4.68 billion, indicating widespread investment in this burgeoning class of weapons. Loitering munitions, also known as “kamikaze drones,” are increasingly used in war zones around the world, most recently in Iran. While these munitions are attractive for militaries because of their low cost, versatility, and relative precision, there are concerns among the international humanitarian community about their use in densely populated areas and the integration of autonomous capabilities.

As a result, our student team at the Human Rights Center’s Investigations Lab began collecting data on the scale and impact of loitering munitions across different conflicts, with a particular attention to their risk to civilian populations. We documented and verified specific incidents to build case studies of loitering munitions, including how they were used, the actors involved, and the harm to civilians. We aimed to highlight how these munitions may violate international humanitarian law as well as gaps in regulatory frameworks and accountability mechanisms.

The impact of a Russian loitering munition on a residential building in the city of Izmail on September 27, 2024. Image by VileGecko via Wikimedia Commons.

What is a loitering munition?

Loitering munitions are a self-destructive subset of unmanned aerial vehicles, or drones, that combine features of traditional surveillance drones and guided missiles. They are often equipped with the cameras and sensors common to reusable surveillance drones, which they use to identify targets as they fly, or “loiter,” over a designated area. Once they identify a target, they use integrated warheads that explode when the drone itself strikes, hence the nicknames “suicide” or “kamikaze” drone.

Though autonomous capabilities are an increasing characteristic of many military systems and are a key feature of loitering munitions, they are mostly unregulated. Their level of autonomy is characterized as either in-the-loop, meaning that a human operator is consistently engaged in identifying a target and making decisions, on-the-loop, meaning that a human is monitoring and can intervene if needed, or out-of-the-loop, which means that, once launched, the weapon identifies and engages targets on its own.

Loitering munitions are significantly cheaper than traditional warheads and very expensive to intercept. Given this and their extreme launch versatility, with different models deployed from backpacks, trucks, ships, or warplanes, they are fundamentally changing how modern wars are waged, and have become ubiquitous worldwide.

Israel, the United States, Turkey, Ukraine, Russia, Iran, Poland, and China are among the countries with active domestic drone production industries, and most modern militaries are either developing their own manufacturing capabilities or acquiring weapons and technologies from allied countries. The strategic and economic implications of the shift towards drone warfare have been extensively discussed in national security circles, but its effects on civilians warrant serious consideration from international bodies governing the laws of war.

Loitering munitions have been in development for decades, but have been increasingly deployed in warzones around the world since 2020, showing up in conflicts in Palestine, Ukraine, Sudan, and Iran.

U.S. Soldiers from various units across Europe assemble a Switchblade 600 during the culminating event of their Loitering Munitions Training at the Grafenwoehr Training Area, Germany, Nov. 4, 2024. This was the first time U.S. Soldiers launched Switchblade 600s, a loitering munition system, in Europe. Image by U.S. Army Photo by Sgt. Cody Nelson via Wikimedia Commons.

Methodology

In this project, we utilized open source investigation methods — which involve collecting and verifying publicly available information from online sources, including social media — to document the use of loitering munitions in armed conflicts, with a particular focus on their deployment against civilian populations. We used the methodology set out by the Berkeley Protocol to collect and verify online content.

We began our investigation by developing profiles on common types of loitering munitions, detailing their countries of origin, technical specifications, and the conflicts in which they are most frequently used. This process allowed us to become familiar with the different kinds of munitions we were studying and to identify their distinguishing characteristics.

After building these profiles, we moved into the discovery phase, searching social media platforms and local news outlets for evidence of loitering munitions being used against civilians. Our searches included keywords such as “loitering munitions,” “kamikaze drones,” and “suicide drones,” combined with regions identified during our profiling. We also consulted databases such as the Open Source Munitions Portal and AirWars, which track weapon use in conflicts by type, location, and casualties. These sources provided a useful foundation for understanding patterns of munition deployment, allowing us to better refine our searches across platforms like X, Instagram, Facebook, and Telegram.

By corroborating incidents and casualty figures with local news coverage and other open source evidence, we built case studies documenting instances of unlawful use of loitering munitions against civilians. This process also involved geolocating incidents using tools like reverse image search and Google Earth Pro to determine the precise strike locations.

Our investigation revealed that many of the targets for these munitions included civilian populations. Munitions regularly struck residential buildings, tents, hospitals, and vehicles, resulting in deaths, injuries, and destruction.

Case Study

We began by inputting “kamikaze drone” as a search term on various social media platforms. We came across this video on Instagram, depicting a loitering munition identified in the caption as a Russian-made Italmas loitering munition crashing into an intersection in the Ukrainian city of Sumy.

Given that we were unaware of the existence of any “Italmas” loitering munition prior to this discovery, we began the verification process to determine if the weapon in this video is in fact an Italmas. We looked up images of the Russian Italmas loitering munition, noting the pointed front, wide wings, and raised edges. We then compared these features with the loitering munition in the video, which appears to match, suggesting that the caption was correct in identifying this weapon as an Italmas.

To find further open source content related to this incident, we used “Italmas” as a search term on other social media platforms, which brought us to this video posted on X on October 21, 2025, the same date as the previous video. The loitering munition has the same features as that of the first video, and other aspects of the background. The neon green sign, the tall post with an orange symbol, and the positioning of the building make it easy to conclude that these videos almost certainly concern the same incident in the same location.

To gather more details about the incident, we searched for media and news reports that mentioned a Russian loitering munition crash in Sumy on the same date, and came across this image on Reuters news wire. The date of this article and the presence of the same neon green sign behind the civilian vehicle indicate that this image is likely also from the same incident.

To determine whether there were any reported casualties, we did a final search for reports or social media content related to the incident. We discovered this post on Facebook, which includes the same video as that in the post on X, but mentions that local authorities reported nine people being injured in the attack.

This article published by RBC-Ukraine corroborates the reports of nine civilians being injured by this attack, and it also includes an image of an orange vehicle that was struck by the weapon, which is the same vehicle that can be seen in the previous video on X, right before the strike occurs.

To verify the location, we relied on another post on X depicting a different Italmas loitering munition strike in the same locale, this time targeting a gas station.

We used Google Maps by inputting the location stated in the posts’ captions, and then locating all gas stations within this district.

To narrow down the options, we determined that, based on news articles’ depictions of the green, white, and red exterior of the gas station that was hit, the gas station in question was likely a Marshal.

To determine which Marshal was the correct one, we investigated the other notable features present in the video. We researched the blue-and-white-striped building in the background, which we determined to be Епіцентр, or an Epicenter store.

Google Maps reveals that there is only one Marshal gas station in the Sumy district that is located next to an Epicenter store.

By accurately positioning Google Maps street view in reference to the major landmarks in the videos — namely the neon green sign and the tall post with an orange symbol — we verified that the location where the video was taken is roughly represented by coordinates 50.91716043217655, 34.83421455863712.

Takeaways

There is a need for internationally legally-binding rules on loitering munitions. Currently, the UN Convention on Certain Conventional Weapons has only produced non-binding guiding principles through a two-tiered framework for autonomous weaponry, including loitering munitions: (1) prohibition and (2) regulation. While fully autonomous weapons that are indiscriminate in impact are prohibited, the qualifications for regulating munitions are vague. In addition, international regulation of loitering munitions is currently opt-in and thus unenforceable.

When examining attacks after the fact, it is difficult to verify whether machines are human-in- or human-out-of-the-loop. The incorporation of AI by companies for military purposes complicates this further. Existing regulations have not addressed AI’s high possibility for coding errors and susceptibility to cyber attacks, both of which heighten risks to civilians. This ambiguity raises serious concerns around plausible deniability, where actors may attempt to avoid accountability for attacks they claim were conducted by weapons controlled by AI.

Additionally, there are concerns about targeting even when operations run smoothly, as there is limited transparency about the datasets that train AI and which are used by these weapons to acquire targets and inform actions. This raises questions within the international humanitarian law community about compliance with the legal principles of target distinction — which protects civilians and civilian objects from being the focus of an attack — military necessity, and proportionality of civilian harm. There are no substantial international laws that require disclosure of whether an attack is conducted by a human or a machine. Burning questions include who ultimately bears responsibility for civilian harm involving autonomous weaponry? Is it the company that developed the weapon? The company that provided and trained its AI components? The operator that launched it?

There is also a significant lack of transparency surrounding the export of these weapons. Loitering munitions produced in one country are subsequently exported around the world and have been used to target civilian populations. The multitude of nations producing and exporting such weaponry and the proliferation of war make the need for trade laws urgent.

As evidenced by our research, loitering munitions are at the frontier of a next generation of warfare that threatens tremendous harm to civilians around the world. International governing bodies and the nations that support them must develop enforceable legal frameworks to address this fragmented accountability landscape.

To learn more about the work of HRC’s student-powered Investigations Lab, visit our website at humanrights.berkeley.edu.