According to Klonowska, one of the most significant developments of the past decade is the use of algorithms to identify objects, including military complexes, tanks and warships, but also schools and hospitals. AI systems are also being used to predict the behaviour of individuals and map their social relationships based on data from social media and intercepted communications. AI is now even being applied to the development of attack strategies. Klonowska stresses that despite the ambitious goals, these systems are technically far less advanced than is commonly suggested.

A central problem is that some militaries place such a high premium on speed that they overlook the limitations of AI. This is compounded by the hype generated by commercial companies, which creates inflated expectations about what the technology can deliver in practice. Many AI applications are deployed before they have been adequately tested, or are used for tasks they were not specifically designed to handle. As a result, accuracy is pushed into the background, with a greater risk of civilian casualties as a direct consequence. Where humans make mistakes at an individual level, AI can do so on a far larger scale due to the speed at which it can generate targets.

Klonowska also highlights the difficult relationship between human oversight and AI systems. People are strongly influenced by AI outputs, making it particularly difficult in practice to intervene at precisely the right moment when a system fails. This is further complicated by what she calls “tinkering”: the continuous adjustment of interfaces, data and models by engineers working alongside military commanders. Every small change in the code can have major consequences, and updated systems can rarely be adequately tested before being redeployed.

Existing legal frameworks offer limited protection. Much regulation requires the taking of precautionary measures, but in high-speed operations there is little practical room for this. Klonowska argues that understanding when AI fails, for example under certain weather conditions, is one of the greatest challenges currently facing the sector. As long as that question remains unanswered, the deployment of AI in military contexts remains a legal and ethical issue of considerable magnitude.