RP4: The Scope of an Autonomous Attack

Kwik J, ‘The Scope of an Autonomous Attack’, in C. Kwan, L. Lindström, D. Giovannelli, K. Podiņš, D. Štrucl, CyCon 2024: Over the Horizon (2024, NATO CCD COE, Tallinn)

Abstract:

‘Attack’ is an important term of art in international humanitarian law
that serves as the basic unit of reference for many targeting obligations. It is often
also asserted that human commanders of autonomous weapon systems (AWS) must
make legal determinations ‘per individual attack’. Divergent interpretations on
what constitutes an attack nevertheless lead to drastically different conclusions with
regard to the technology’s lawfulness: interpreted narrowly (‘each shot’), it precludes
AWS technology entirely, while interpreted broadly (‘each activation’), it sanctions
extensive autonomous activity. This paper theorizes that imprecision on the scope of
attack is an underappreciated aspect of the AWS controversy that hampers theoretical
and diplomatic advancements. The legal boundaries of autonomous attacks are
analysed through the lens of targeting law, and a scaling methodology is proposed that
allows commanders to determine the maximum extent to which autonomous activity
may still lawfully be grouped into one single attack. The paper argues that both overly
narrow and broad interpretations are inconsistent with targeting principles and practice,
instead favouring a middle-ground approach based on temporal and spatial proximity
that properly respects international humanitarian law’s (IHL) balancing philosophy
between humanitarian and military interests. Through consideration of practical
scenarios, the paper subsequently demonstrates how this impacts the application of
targeting rules, such as at what intervals the commander’s duty to verify or cancel is
triggered and under what circumstances successive autonomous engagements may be
grouped together for proportionality assessments.

Keywords: attack, targeting, autonomous weapons, precautions, IHL
proportionality, proximity in time and space