Originally published 2021-07-02.

In mid-2021 several outlets, led by New Scientist, reported that the Israel Defense Forces had used a swarm of drones guided by artificial intelligence to find and strike targets in Gaza during the 11-day war with Hamas in May 2021. If accurate, it would be one of the first known combat uses of a coordinated drone swarm. The Israeli military has spoken openly about using AI to process intelligence, but it has not published details of how the swarm worked or how much of it was automated.
That gap matters. "AI drone swarm" can mean anything from a group of reconnaissance drones sharing data to weapons that choose and attack targets on their own. This article sets out what was reported, what remains unconfirmed and why experts worry about where the technology is heading.
What was reported about Gaza in May 2021
The fighting began on 10 May 2021 and ended with a ceasefire on 21 May. According to IFLScience, more than 4,400 rockets were fired into Israel and around 1,500 Israeli strikes hit Gaza in those 11 days, with hundreds of people killed, most of them in Gaza.
The reports, which appeared at the end of June and early July 2021, said a networked group of drones was used to locate, identify and help strike militants and rocket launch sites. Secondary write-ups described a swarm that could keep working if some of its drones were lost, drawing on data from satellites, other aircraft and ground intelligence. Fox News, summarising the claim on 5 July 2021, noted that it rested on the New Scientist report and included no official confirmation.
What the Israeli military has said about AI
The IDF has described itself as using artificial intelligence and supercomputing heavily for about two years before the 2021 war. Its stated aim was to sift huge amounts of information faster, pinpoint suspected Hamas positions and missile launch sites, and shorten the fighting.
Much of this work is linked to Unit 8200, the Intelligence Corps unit best known for signals intelligence and code-breaking. Its algorithms were reportedly used to combine geographic, human and signals intelligence into lists of potential targets, which were then passed to commanders who ordered the strikes. In other words, the AI described in official comments is mainly a tool for analysis and targeting recommendations, with people still in the decision chain.
The IDF did not release specifics about the swarm itself, so claims that it was "fully AI-controlled" go further than the available evidence supports.
Swarm, autonomous weapon or decision aid?
It helps to separate three ideas that are often mixed together in headlines:
- Drone swarms are groups of drones that coordinate their movements with each other, often with software handling the flying so that one operator can manage many aircraft.
- Autonomous weapons are systems that can select and attack a target without a human approving that specific strike.
- AI decision support means software that analyses intelligence and suggests targets, while humans decide whether to attack.
A swarm can exist without autonomous killing, and AI targeting can exist without any swarm. Based on what was made public, the Gaza case involved a coordinated swarm and AI-assisted targeting, but whether any drone fired without a human decision has not been established. Arthur Holland Michel of the UN Institute for Disarmament Research told IFLScience that such systems "fall quite far short of the large dynamic, intelligent swarms" that could one day change warfare.
The earlier Kargu-2 case in Libya
The Gaza reports came a few months after a UN Panel of Experts report on Libya, published in March 2021. It said that in 2020 a Turkish-made STM Kargu-2 drone had "hunted down and remotely engaged" retreating fighters loyal to General Khalifa Haftar, describing the weapons as having a "fire, forget and find" capability. Some commentators called it the first attack by an autonomous drone.
That claim also has limits. The report did not say clearly that the drone acted after losing contact with an operator, nor that anyone was killed. STM's own promotional videos show a human choosing the target before the drone dives on it. The Kargu was also reportedly used by Azerbaijan in the 2020 Nagorno-Karabakh war, but that has not been confirmed by Azerbaijani authorities.
Why experts are worried
Zachary Kallenborn, a research affiliate at the University of Maryland's National Consortium for the Study of Terrorism and Responses to Terrorism, studies autonomous weapons. Speaking to Live Science in June 2021 about the Libya case, he explained that AI systems only recognise what they were trained on, so messy real-world situations can fool them. "A soldier might wear camo, or a farmer a rake, but a farmer might wear camo too," he said.
He was even more concerned about swarms. "In my view, autonomous drone swarms that can kill people are potentially weapons of mass destruction," Kallenborn told the magazine, because large numbers of cheap, coordinated drones could cause harm on a scale once reserved for far bigger weapons.
Critics raise several other problems:
- Errors and bias: a system trained on incomplete data may misidentify civilians as fighters, and mistakes can repeat at machine speed.
- Accountability: when an algorithm contributes to a deadly mistake, it is harder to say who is legally and morally responsible.
- Distance from killing: automating parts of the process may lower the psychological barrier to using force.
- International law: the laws of war require distinction between combatants and civilians and proportionality in attacks, judgments that are difficult to encode.
Supporters argue the opposite case: that faster, more precise analysis could reduce mistakes and shorten wars. The debate is unresolved, and the evidence from real conflicts is still thin and largely controlled by the militaries involved.
Efforts to regulate killer robots
Groups such as Human Rights Watch and the Campaign to Stop Killer Robots have called for a pre-emptive ban on fully autonomous weapons, arguing that decisions over life and death must stay with humans. Governments have discussed limits for years at the United Nations under the Convention on Certain Conventional Weapons, but no binding treaty has been agreed. Many states support the principle of "meaningful human control" over the use of force, while major military powers have resisted firm restrictions.
For readers trying to judge new headlines about "AI wars", a few questions help: Who confirmed the claim? Was the system choosing targets, or helping humans choose them? Was a human required to approve each strike? Clear answers to those questions are still rare, which is exactly why careful reporting matters.
Sources: IFLScience, 2 July 2021: Israel used world's first AI-guided swarm of combat drones in Gaza attacks; Fox News, 5 July 2021: Israel used swarm of drones, report says; Interesting Engineering, 5 July 2021; Live Science, 3 June 2021: interview with Zachary Kallenborn on the Libya drone attack; Wikipedia: STM Kargu. The reports draw on New Scientist's original story of June 2021.

COMMENTS