Designing compliance

This PhD project examines the role of the private companies designing and developing military AI capabilities for states, particularly decision-support systems. It seeks to conceptualise, theorise and analyse the role being played by a non-state actor within the traditionally state-centric discipline of international law, drawing from empirical insights on how public and private actors are collaborating together throughout the AI lifecycle and their decision-making pertaining to (legal) compliance.

AI, the private sector and state compliance

Increasingly, open-source intelligence and big data are providing actionable military insights for the targeting operations by armed forces. To this end, the private sector is developing AI-powered software that collects, collates, and analyses data for military intelligence (AI-decision support systems). Though there are different arguments as to whether such AI tools will improve or be a risk to international humanitarian law (IHL) compliance, it is also invariably said that IHL rules can be embedded into the code of this software, such as to have compliance “by design”. At the same time, there are also (presently) non-binding “responsible AI” requirements at the level of attaining a particular standard by the technological system. Taken together, the interaction of these legal frameworks with machine-learning based AI may shape what key IHL targeting rules mean and entail in practice due to the quantification logics of algorithms and human-machine interaction.

While states have the ultimate responsibility to comply with international legal rules, ex-ante decision making by private companies is increasingly important in influencing the conduct of state actors at the point of end use. As such, this PhD project aims to analyse the private sector’s role in relation to state compliance against the above complex and oftentimes vague regulatory framework, and consider this from the perspective of how public and private actors are collaborating throughout the AI lifecycle to uphold end-use compliance.