Drones are in the news again, but not as we have come to know them. In August, Britain announced it was sending 850 Black Hornet “microdrones” to Kyiv for use in close-quarters combat.

The idea was (before the spectacular Russian collapse in recent weeks) that they would lend Ukrainian troops a crucial edge in the vicious urban fighting that was expected as they sought to liberate their towns and cities.

These machines are a far cry from the large unmanned aerial vehicles (UAVs) associated with the war on terror, the ubiquitous Predator and Reaper drones that delivered death from the upper skies with almost god-like insouciance.

Black Hornets are actually more like a child’s toy. Measuring just over six inches and weighing a little less than a plum, they will literally peer round corners and sneak through windows.

Due to their minimal rota noise, they can creep right up on the enemy, entering Russian compounds and defensive positions (any that are left) and beam back high-definition footage and stills via three cameras mounted in the nose.

The Hornet is piloted by a frontline soldier via a small iPad-style screen and a hand controller that would not look out of place on a 1990’s games console. At £10,000 a pop, but bound to get cheaper, it offers what soldiers since before Thermopylae have craved: situational awareness and the ability to “see over the hill” – without having to send some poor blighter to the top of it.

Dominic Nicholls, The Telegraph’s Associate Editor, knows only too well the value of a remote pair of eyes. He used drones, big and small, on numerous operations, including in Iraq and Afghanistan.

“The drones I worked with allowed us to make decisions based on the best available real-time information,” he says. “That might have been selecting the best helicopter drop-off points for soldiers assaulting a position, or talking other assets on to potential threats.”

Just as in civilian life we will shortly be using small drones to deliver our packages, militaries are now finding uses for them that a few short decades ago would have been written off as science fiction.

Ukraine in particular offers a very public demonstration of their versatility – and a glimpse of the future.

The first major war to be played out in real-time on social media, the conflict has sparked a small revolution in drone innovation, with commercial and domestic models – many of them donated by the West – hacked and modified to deliver lethal force.

It is now commonplace to go onto Twitter and see footage of an adapted drone, which last year might have filmed your wedding, dropping a grenade onto a bivouac of sleeping Russian infantry, or through the hatch of a tank. Intelligence suggests this has been somewhat bad for morale. The invaders feel hunted, spending as much time looking up as forward.

In short, drones are having a very good war. They are clearly the future. And if the machines can help technologically advanced nations like the West and its allies win battles while sparing their troops, what’s not to like?

Well, if a growing movement of really quite eminent scientists is anything to go by, rather a lot. What, they ask, if the machines are becoming too effective? 

For the apparently positive example of Ukraine, they might well counter with those of Libya and Gaza. That’s because these are locations where – amid all the usual horrors – small drones are recently suspected not only to have killed people, but to have done so autonomously, independent of human control. 

Minor blips on the radar, you could argue. But across the world a moral line is being crossed. We are entering the much prophesied age of the killer robot – a march put on steroids by the availability of increasingly cheap and effective unmanned flying machines and the artificial intelligence driven by the demands of the smartphone.

The question is, do we know what we unleashed? And, by the time we’ve worked it out, will it be too late?

To illustrate the point, it is worth understanding what systems like Black Hornet cannot do. 

Announcing the UK’s aid package, Defence Secretary Ben Wallace described them as “cutting-edge”. But even a lowly computer science graduate student would roll their eyes at that. Why? Because it ignores the fact that the Hornet still requires a soldier to put him or herself in danger by getting close enough to the enemy (less than 2km) to pilot it into position, and that once there the drone can only act passively, as a reconnaissance tool. In short, it cannot think for itself and it cannot kill.

But that is changing rapidly.

In 2017 a short film was premiered at a United Nations diplomatic conference in Geneva which, to put it mildly, ruffled some feathers. Slaughterbots – an award-winning short film from the Future of Life Institute – portrayed a fictionalised slick, Steve Jobs-style tech presentation introducing a microdrone small enough to fit in the palm of your hand.

The audience laughed and clapped as the machine playfully dodged the presenter’s attempts to grab it. Even when he didn’t try, it seemed to dance in the air, apparently displaying its “stochastic motion” anti-sniper function. 

The best part? “It’s all AI,” the presenter says smugly, in the film. “Its processor can react 100 times faster than a human.” In other words, it controls itself. He goes on: “Just like any mobile device these days, it has cameras and sensors, and just like your phones and social media apps, it does facial recognition.”

He then reveals that inside the tiny four-rotored drone, which has by now landed gently in his hand, is three grams of “shaped” explosive.

He throws it into the air, whereupon it swiftly recovers its orientation and charges like a bolt of lightning at a nearby mannequin, blowing a small hole in its forehead. “That little bang,” we are told, “is enough to penetrate the skull and destroy the contents.”

Pretty terrifying: a small, inexpensive drone using facial recognition to select and kill a target, free from any direct human control.

Slaughterbots, of course, is not a documentary. But how close are we to the real thing?

Stuart Russell is professor of computer science at the University of California, and was a sponsor of the 2017 film, which has now been viewed more than 75 million times.

Perhaps it’s his measured British accent, or the box of PG Tips visible in his Berkley office as we talk via Zoom, but he doesn’t come across as a common or garden doom monger. In fact, he is one of the world’s leading brains in this field and delivered last year’s Reith Lectures on the subject of living with artificial intelligence.

The key point, he says, is that the technologies required for a lethal microdrone of the type portrayed in the film already exist. They just need to be improved in some areas, shrunk down and integrated.

“There are some technical issues in miniaturisation which have to do with batteries and range and how much onboard power you need to support the computer processing you need to have a fully autonomous system,” he said. “But really that’s an engineering task. There’s no science fiction stuff that needs to happen. On most of that we’re already there.”

To build a truly high-performing microdrone, you would probably design your own lighter and more power-efficient computer chip, known as an Application-Specific Integrated Circuit (Asic), rather than using a general purpose chip. But that’s perfectly doable.

The fictional Slaughterbots presenter said it himself: a lot of the tech that would enable that microdrone to find you and then decide whether or not to kill you – cameras, facial recognition, geolocation – already exists in your smartphone.

Had the film ended there it would have been unsettling enough. But it goes on to show the devices operating as part of a pack, or “swarm”, hunting down a bunch of supposed bad guys. And this is really the whole point of the armed microdrone concept. On its own, one could be very nasty, but as part of a coordinated group, an absolute nightmare.

AI computer systems could be “trained” to operate multiple drones as a team, penetrating buildings, trains, cars, pretty much immune to bullets and less vulnerable to electronic jamming because they’re not being controlled remotely.

Just picture it: a couple of the team attach themselves to an air duct on the side of a building, blow themselves up, and the rest swarm in through the filtration system, perhaps programmed with facial, iris or even gait recognition software.

Dominic Nicholls imagines their use in war. “Once let loose across enemy lines to hunt for specific targets, no senior commander, no matter how close to the action, will ever feel any measure of security on the battlefield.”

Again, we are not there yet, but if this scenario is futuristic, then it’s the near future. 

For years now, drones have been working autonomously in teams for tasks like search and rescue and cleaning up oil spills. And the first signals of their use for lethal force are beginning to emerge.

In 2020 a skirmish in Libya involved the possible use of a Turkish-made autonomous lethal drone called Kargu-2. According to a UN report, this was used to hunt down and engage a logistics convoy affiliated to the warlord Khalifa Haftar, marking the known battlefield debut for that class of weapon.

Last year, meanwhile, it was reported that the Israel Defence Forces deployed a swarm of drones to locate, identify and attack Hamas militants, thought to be the first time a swarm has been used in combat. 

None of these strictly speaking were microdrones. But the proof of concept, particularly in the case of the Gaza incident, is clear enough.

“So what?” you might ask. After all, the history of warfare is in one sense defined by the introduction of new and unpleasant weapons that make it easier to kill the enemy. Aren’t microdrones merely the latest?

Yes and no. The crucial factor, according to Professor Russell, is “scalability” – not a particularly sexy word, but one with potentially era-defining implications.

Let’s return to Slaughterbots, the short film he backed that has gained such currency among LAWS (Lethal Autonomous Weapons Systems) sceptics. In its second act we leave the imagined tech presentation theatre (with its deceased mannequin) and pan instead to a large military cargo plane in flight.

“We are thinking big,” the voiceover states. “A $25 million order now buys this…Enough to kill half a city – the bad half.”

With that the rear door of the aircraft opens and thousands of microdrones tumble out, activate in the air and begin buzzing towards the ground like malevolent wasps. This is a swarm, but on a different scale. The microdrone has effectively become a weapon of mass destruction.

If that sounds over the top, consider that only a few years ago the Black Hornet cost around $80,000 a unit; now they’re an eighth of that price.

It is astronomically difficult to build your own nuclear weapon, or even to buy one, if the international community disapproves. Just ask Iran. But as autonomous drone technology proliferates as an increasingly mainstream aspect of commercial and recreational life, the same will not be true of swarms.

How hard would it really be for a malign state to set up a super factory that pumps out hundreds of microdrones a week, or a terrorist group like Isis or Al-Qaeda simply to buy the machines on the open market?

Most observers believe – albeit perhaps wishfully – that Vladimir Putin will not resort to a nuclear attack for fear of the international response. But would the man who shows no compunction in flattening civilian areas with artillery hesitate to use a mega swarm of killer microdrones to re-take Kharkiv or Izium?

Programming drones to attack gun-carrying humans in military uniforms is technological child’s play. But why stop there? Why not also automatically track mobile phone use and take out civilians who have posted pro-Kyiv content on Twitter or Instagram?

This kind of integrated capability may still be years away; military procurement (the F-35 fighter, for example) is normally measured in decades. Equally, it could be much closer.

Professor Russell says there are clear parallels with the creation of the atom bomb. By the end of the 1930s, much of the theoretical heavy lifting, such as identifying which uranium and plutonium isotopes to enrich, was in place.

But a big push was needed to engineer this into a working reality. In 1943 the US government opened its chequebook, gathered the nation’s finest scientists in Los Alamos, New Mexico and diverted one per cent of the country’s entire electricity supply to their endeavours. Two years later they had their bomb.

“It’s a matter of will,” he says. “If we wanted to create that kind of weapon, I think we could do it in 18 months. It’s easier than a Manhattan project.”

These concerns are not confined to the university campus.

Chris Cole is a softly spoken, thoughtful kind of chap. He runs a website called Drone Wars UK from his home in Oxford which aims to document and warn against the use of armed drones, and particularly British involvement. 

“There’s an ethical issue when machines decide who to kill,” he said. “That line is being corroded all the time. We are getting to the stage when humans are rubber-stamping decisions being made by machines.”

Mr Cole and his team spend much of their time organising public meetings and mass letter-writing campaigns, as well as occasionally taking to law to challenge alleged government secrecy on the subject.

If that sounds familiar, perhaps it’s because his organisation has firm links with the Campaign for Nuclear Disarmament and various anti-arms trade bodies. But that in itself is revealing: the potential for nuclear armageddon remains, but its longtime opponents now appear to view the rise of small plastic robots as a comparable threat.

Might we be entering a new arms race similar to that of the 1940s and 50s?

The west is in a quandary. Its leading military powers clearly view autonomous weapons as the future of warfare. But they are understandably sensitive about public reaction to the idea of fully-independent killer robots. Many of us grew up on the Terminator movies, after all.

The US Department of Defence has since 2012 operated under self-imposed Directive 3000.09. It’s vague. Some understand it to mean that a human has to be present in the decision-making “loop”, although others disagree, or say this stipulation is easily overridable anyway.

Look carefully and the direction of travel appears plain. Announcing an update to its CODE (Collaborative Operations in Denied Environment) program some years ago, a DARPA (Defence Advanced Research Projects Agency) official said: “Just as wolves hunt in coordinated packs with minimal communication, multiple CODE-enabled unmanned aircraft would collaborate to find, track, identify and engage targets, all under the command of a single human mission supervisor.”

The fundamental nature of battle in the middle part of this century and beyond depends upon exactly what that supervision entails.

Some countries see the direction of travel and want it to halt: since 2014 dozens have signalled their support for a treaty banning the use of autonomous killer robots. Others want to get ahead.

These are big players. Neither the US, China, India, nor Russia has arguably shown much real willingness to sign up. Nor, for that matter, has Britain. This year the RAF killed off its Project Mosquito drone (swarm) program, under the auspices of 216 Squadron, but only in order to move forward with plans for a better replacement.

“This is one of the few areas where the US, UK and Russia are all in agreement: that we should create this new class of weapons,” Professor Russell chuckles ruefully (needless to say, he supports a ban). His frustration is evident. “I think the UK could have a significant role to play if it were to engage seriously with the issue. The window to act is closing fast.”

Approached ahead of this article, the Ministry of Defence denied that the UK possessed fully autonomous weapons systems, i.e. those without “context-appropriate” human involvement, or that it intended to use them.

It said its strategy was to promote “safe and responsible” military development of AI in concert with allies and push for compliance with current international law, which it described as “highly effective”.

Perhaps, as with nuclear, there is no halting the march of technology. And is it not natural, indeed humane, to seek ways to keep soldiers out of the line of fire?

Major General Jonathan Shaw has seen the bloody cost of conflict up close, first as a platoon commander with the Parachute Regiment in the Falklands, later commanding a multi-national division in southern Iraq.

“Why are people interested in the air?” he asks. “It’s because in the army we’re always asking ourselves ‘what’s over the hill’?” Microdrones, he says, “allow you to fight a war without human costs.”

Although not expressing a view on the question of automation, he said that with drones “you’re killing more of the enemy without any extra cost to yourself, and that’s always the goal. They would have been really useful in the Falklands. There were numerous casualties.”

As a simple statement of fact, which is no more than the General intended, it’s difficult to argue with. But perhaps this view begs a certain question.

Artificial Intelligence – robots, in other words – are demonstrably better than humans at a large and growing range of activities, from playing chess to piloting aircraft. But, if you take into account the crucial moral factor, can they ever be better at deciding who gets to live and who should die?

Remember, the introduction of self-driving cars is currently snagged on just this dilemma.

For all their many messes and controversies, the Predator and Reaper drone strikes that we became so familiar with after 9/11 were, as far we know, fully controlled, not just “supervised”, by people. 

“There’s no euphoria after a drone strike,” says Dominic Nicholls. “I’d describe the overwhelming emotion as one of relief. You have to trust the intelligence that has taken you to the point of deciding to use lethal force against an individual. But once that decision is taken you want the act to be carried out in as efficient a manner as possible. 

“I never saw anyone punch the air or in any way take delight in the violent death we had just brought about. Most people, if they expressed any thoughts at all, were measured, professional and glad all parts of the system had worked.”

It’s not perfect. But it is human.

Protect yourself and your family by learning more about Global Health Security