Originally published 31 March 2022.

The Russian invasion of Ukraine and the poorer than expected performance of the Russian army have prompted fierce debate among military commentators on why Russia’s much-vaunted military reforms of the past decade—particularly the integration of artificial intelligence technologies that were supposed to enhance Russia’s joint operations capability—seem to have been unsuccessful.

So far, Russia’s deployment in Ukraine has been a demonstration of some of the limitations and vulnerabilities of AI-enabled systems. It has also exposed some longer-term strategic weaknesses in Russia’s development of AI for military and economic purposes.

Russia’s use of AI-enabled technologies in the invasion reportedly includes disinformation operations ^[1], deep fakes ^[2] and open-source intelligence gathering ^[3]. But information operations are not the sum total of Russia’s AI capabilities. AI is embedded across the military spectrum, from information management, training, logistics, maintenance and manufacturing, to early warning and air-defence systems.

Since at least 2014, Russia has deployed multiple aerial, ground and maritime uncrewed systems and robotic platforms, electronic warfare systems, and new and experimental ^[4] weapons in both Syria and Ukraine.

The AI elements in these systems include image recognition and image stitching in Orion combat drones ^[5], radio signal recognition in Pantsir-S1 anti-aircraft systems, AI-enabled situational understanding and jamming capability ^[6]in the Bylina ^[7] electronic warfare system, and navigation support in the Kamaz truck ^[8]. So-called kamikaze drones ^[9] (developed by the Kalashnikov Group, the maker of the famous assault weapon) appear to use a mix of manual and automated ^[10] target acquisition.

One of the earliest images that circulated online ^[11]in the current conflict was of a Russian Pantsir-S1 stuck in the mud in a field in southern Ukraine. The Pantsir is a component of the early warning and air-defence system that features both short-range surface-to-air missiles and 30-millimetre automatic cannon.

If we look under the hood at the purported AI technologies of the Pantsir, Russian state media Izvestia reported ^[12] two years ago that it

is capable of detecting, classifying and firing at air targets without the participation of an operator. The developed algorithms instantly determine the importance of objects and arrange the order of their destruction depending on the danger they represent …

Its software takes into account the tactical situation, the location of targets, their degree of danger, and other parameters and selects the optimal tactics for repelling a raid.

It’s wise to be wary of Russian claims of full autonomy, both because the capacities of AI-enabled systems tend to be overblown and because deliberately fabricated information ^[13] is a component of Russia’s approach to new-generation warfare ^[14].

There are reasons to be concerned about the individual systems and their AI components, but it’s important to consider Russia’s grander AI vision.

The Pantsir-S1 is just one node of an interconnected system that includes airborne radar systems, satellites and reconnaissance drones, and panoptic information-management systems.

Just as the West has been pursuing a vision of an interconnected battlefield in the form of a ‘joint all-domain command and control centre’ (JADC2) concept, Russia has its ‘national defence management centre’ (NDMC), which aims for the same. The goal is to build systems in which ‘data can move seamlessly ^[15] between air, land, maritime, space, and cyber forces in real time’.

According to researchers at CNA, a US think tank, ‘NDMC was designed ^[16]to receive information from the lowest military unit levels, and, following analysis and evaluation, feed the data directly to those at the strategic level.’ This defines the battlefield in multiple dimensions and makes shared situational understanding contingent on information from the edge ^[17] of the battlespace.

There has been some speculation that poor-quality tyres ^[18] or failure to account for local ^[19]conditions indicate vulnerabilities in Russian forward planning. While these physical aspects are important, in a battlespace that’s dependent on information it’s also important to consider the extent of interoperability of systems, any bandwidth constraints, and the tendency in computer-assisted decision-making to equate a map with territory. Expanding a military’s capacity to use so-called real-time information requires analyst teams to interpret and leadership to prioritise the information.

The critical human element was hindered because Russia’s planning seems to have been tightly held within President Vladimir Putin’s inner circle (which does not include the army) until just before the invasion.

Putin said in 2017 ^[20], ‘Whoever becomes the leader in this sphere [artificial intelligence] will become the ruler of the world.’ Since then, he has pushed the use of defence sector spending and defence acquisition ^[21] to generate national economic growth and drive national technological innovation.

With few reasons to adhere to international norms on ethical AI development, including regulation of data harvesting, as well as a ready supply of programming talent, Russia could be perhaps seen as having some advantages in AI developments, and the ability to quickly deploy innovations into its asymmetric warfare programs.

However, the two directions ^[22] in which the Kremlin is steering ^[23] the Russian defence sector—increasing civilian and dual-use goods and import substitution—carry some distinct vulnerabilities for Russia. Despite pushes by Putin in recent years towards economic self-sufficiency, the interconnected nature of global trade means that there are key technology choke points that affect Russia’s AI development.

The main barrier to Moscow’s vision of AI supremacy is microchips. Russian media has claimed ^[24] that NDMC’s information management runs on ‘Russian-made’ Elbrus microprocessors ^[25]. However, Russia lacks the ability to produce these microchips. The production of Elbrus chips is outsourced to Taiwan, to the company TSMC, which has now suspended ^[26] its production and export to Russia.

At a broader level, at least 1,300 Russian defence enterprises ^[27] have shortages in human capital, particularly in process engineering, and the exodus of tech talent from Russia ^[28] has been accelerated by the war. A range of Western sanctions in key technologies and commodities will hit the Russian economy hard ^[29], even if it tries to substitute with imports from China.

Some analysts ^[30] have claimed that Russian underinvestment in technology modernisation has hampered the army’s ability to ‘see’ the battlefield, forcing it to rely on the brute force of tanks and artillery. For those Ukrainians now under siege in cities around the country, the apparent modernisation of Russia’s war apparatus will be perhaps a moot point as artillery rains down. But AI is a critical technology ^[31] that will be increasingly important for both economic development and national defence. It’s important to understand both the way it is embedded in technologies and the international supply-chain interdependencies that are crucial for its development.