Online Conference 5-11 July 2021
February 2–7, 2027
Vienna, Austria
Parallel Papers Session
Day and time: to be announced
Paper code: PP27K
Presenter: Biljana Stefanovic
Moderator: to be announced
Biljana Stefanovic
Biljana Stefanovic is an advisor at the intersection of digital transformation and organisational life, based in Belgrade. She has spent her career close to the places where technology strategy is actually decided, in national AI strategy and in industry, and pairs deep enterprise architecture and data & AI expertise with training in systems-psychodynamic organisational consulting. She is drawn to what stays hidden beneath transformation programmes: how organisations and nations meet dependency, uncertainty and the wish to be in control.
Governments and enterprises increasingly demand to own the whole of artificial intelligence - the chips, the compute, the models, the data, the ground beneath them. "Sovereign AI" is the name for this demand. It is a demand to be master in one's own house: to draw a boundary, pull the capability inside it, and govern what is within by one's own authority. This paper argues that the demand misreads its own situation, and that the misreading is revealing.
The argument begins not with anxiety but with structure. Sovereignty assumes an environment - an outside that a state or firm faces across a boundary and might, with enough investment, bring within. But artificial intelligence is not an environment of that kind. Its compute, models and data do not sit outside the actor, waiting to be internalised; they constitute the actor's capability from the start, through relationships that cross every boundary and reach through the organisation itself. The outside, in the sense sovereignty needs, does not exist. What an actor can govern within its own edge is not where the decisive risks lie. Those arise from how it is positioned within relationships it does not control and cannot pull inside. Sovereign AI is the attempt to convert this outer, relational exposure into something internal and ownable. The attempt cannot succeed, because the exposure is constitutive, not incidental - which is why the boundary must be drawn again and again, and never holds.
Seen this way, the same structure appears in a government/state and in a corporation, not by analogy but identically: both are organisations turned inside-out, required to follow actors whose demands they had assumed would follow them, reaching for a supply-side vocabulary of ownership in a world that has become demand-led and asymmetric. The tension between West and East reads less as two blocs each pursuing self-sufficiency than as everyone caught in the same asymmetric dependence, reaching for the same obsolete word.
Where, then, does the anxiety come in? Here an older insight holds, but in a new place. To face the inside-out condition - to accept that one's capability is constituted by relationships one cannot master - is hard to bear. Holding the boundary as a fantasy is easier. The sovereignty programme, with its roadmaps, national models and datacentres, functions as a social defence: it binds the anxiety of a boundary that will not hold by producing ownable objects to point at. The defence does not cause the misreading; it manages it.
The paper ends on a question about authority. If there is no inside to the house - if a country or company cannot pull AI within its own walls and govern it as its own - then what does leadership even mean? It would have to mean leading across a line one does not control, working through relationships of dependence rather than commanding a territory one owns. Yet those who hold authority rarely let themselves consider this, treating it not as the real work before them but as a danger to be kept out - because to accept it is to accept one is not master in one's own house. I write from inside these systems, in national strategy and in industry. The paper does not resolve its question, holding that to resolve it would repeat the omnipotence it examines.