New Scientist
Ukraine’s battlefield is producing something almost as valuable as weapons: vast quantities of real-world combat data for training military AI.
The scramble for Ukraine war data to feed AI
Author: Matthew Sparkes
GLOBAL superpowers and giant technology firms are vying for real-world data from the war in Ukraine to build next-generation AI models that will empower militaries in future conflicts. Videos of drone strikes and other sensor data are like “gold dust” for training AI, experts say.
Last month, Ukraine signed a deal with the UK to share a vast treasure trove of data collected from drones, cameras and sensors since Russia staged its full-scale invasion in 2022. It is claimed the deal will make the UK the first foreign state to have access to Ukraine’s so-called Avengers AI model, which is already being used to identify adversaries and their equipment in real time and slash the time taken to stage attacks.
Access to actual, real, credible battlefield data from Ukraine is a big commercial advantage
The UK claims that the first fruits of this deal are about to be tested through a project that is turning buried fibre-optic cables into sensors to detect hostile intrusions at an unnamed UK defence site.
Peter Lee at the University of Portsmouth, UK, says the varied data will be invaluable in creating a wide range of defence and military tools, which will make both the UK and Ukraine more secure.
“This is the gold dust that allows the UK to enter the autonomous age of drones,” he says. “You can’t really put a price on the data they’re going to get access to. Add up everything that the UK has given to Ukraine over the last several years – this data is worth all of that and more.”
Ukraine flies thousands of drones a day, both in the war zone and penetrating deep into Russia, and has a vast array of sensors on the front line. This data has long been fed into a government department known as Brave1, which analyses it with AI to find strategic weaknesses and opportunities. It feeds that information back to troops on the front line and defence contractors.
AI models can be used to autonomously steer drones to their targets, monitor thousands of video feeds and other sensors for tell-tale signs of adversaries’ movements, or analyse footage of attacks to spot design flaws in military hardware. The models can even plan strikes and pick targets.
Numerous efforts are under way in Ukraine to use AI to eke out an advantage, including at the defence ministry’s “A1” organisation – which the Ukrainian government launched with UK government support in March, long before this latest announcement.
Lee says the possibilities are so compelling that it is almost certain that China is doing the same with Russian data, taken from the opposite side of the front line – and it probably had a head start. He thinks that, in the earliest stage of negotiations between Russia and China about military and intelligence support for the invasion, China will have raised the issue of data sharing.
War machines
“China understands that data is king when it comes to AI, and that the more data you have, the more powerful your systems are likely to be,” says Lee. “It would be a strategic oversight of grotesque proportions [if China hadn’t sought data].”
Joe Burton at Lancaster University in the UK says that despite the UK’s support of Ukraine with weapons, training and intelligence, it is unlikely that raw data has been shared on a wide scale until recently. But a deterioration in relations with the US may be an incentive for Ukraine to create stronger ties with other allies.
“The Trump presidency has been very volatile in its approach to Ukraine and, if you’re in [Kyiv], it makes sense to have a broader set of partnerships around AI data, which can help support you in what will probably still be a long-term war,” says Burton. “I would not be surprised if Ukraine are hedging their bets.”
Big technology firms are also getting involved. Ukraine has long worked with AI company Palantir, and signed a data-sharing deal with the firm months before the agreement with the UK was announced. US company Clearview AI’s facial-recognition technology has been used to find Russian spies operating in Ukraine, while another US firm, Scale AI, has used satellite data to identify Russian troop movements and detect building damage across Ukraine.
“All these companies are competing for data. If you have access to actual, real, credible battlefield data from Ukraine instead of simulated data that you’re generating in a lab, that’s a big commercial advantage,” says Burton.
In short, if you are pitching your product to states looking for AI tools to protect national security or support their militaries, then having a battle-tested product is a strong selling point.
But despite the vast expansion of data collection, AI analysis and high-tech weapons systems, the day-to-day reality for troops in Ukraine remains a muddy trench, not dissimilar to those dug in the first world war. “This idea that you can deploy a load of robots to the battlefield and you don’t need to send soldiers into harm’s way, it’s not [happening] in Ukraine,” says Burton.
Credits: TCA, LLC.