Description
Private technology and AI companies have become central actors in contemporary security innovation, yet their influence on the legal frameworks that govern war remains poorly understood. This paper examines how the integration of AI-based platform infrastructures such as Palantir’s Maven for NATO redistributes authority over the construction of legal categories in military targeting. Rather than merely supporting decision-makers, these systems classify, prioritize, and transform data into actionable designations of “combatant” or “civilian.” In doing so, they perform legal work upstream, embedding assumptions about who may be killed and on what evidentiary basis into code, data models, and interface defaults. This often occurs outside traditional legal and democratic oversight, allowing technology companies to quietly participate in norm-shaping processes. Using the notion of infra-legalities, the paper traces how mundane technical choices (data interoperability, risk thresholds, labeling schemes) materialize as operational norms of distinction within NATO’s targeting cycle. The argument challenges prevalent regulatory approaches that treat law as external to AI technology and instead shows how legal meaning is enacted within sociotechnical infrastructures. Ultimately, the paper contends that safeguarding international humanitarian law’s protective function requires attending to the privatized sites where legal categories are increasingly made.
