There are two sources of potential disaggregation of agency from territorial jurisdiction in the AI agent-enabled world. First, distributed networks of AI compute which give rise to autonomous AI agents acting in the world, but which cannot be pinpointed to a single location, whether a data center or processing chip. Second, autonomous AI-driven robots which function independently, are mobile and are able to autonomous conduct self-servicing and maintenance, possibly cooperating with other AI robots.
The question is whether, in a hypothetical world where embodied human agency is completely displaced by AI agency, spatiality as the primary meta-ordering principle of life breaks down and the territorial sovereign state is replaced by various sovereignities.
In the first case of distributed AI network processing, the sovereign authority is situated over points of leverage or control over the production and processing network through which agency is produced. Of course, each node of the production and processing network may still be traceable to a physical location, such as the location of a factory, a data center, or a processing chip, and rules of attribution that give rise to ownership over these nodes will confer jurisdiction on the legal owners of these assets, whereupon attributions of responsibility could be made. This is analogous to intellectual property rights of IP assets. However, the distributed nature of such processing units will require a large degree of cooperation among states to jointly coordinate recognition of jurisdiction and enforcement, or alternatively a restriction of global data and internet flows in order to restrict the unwanted influences of external AI agency on domestic systems.
The second type of AI agency in decentralised edge AI robotics and computing pose greater challenges, and is already underway with the development of Nvidia RTX chips, JEPA, edge AI and world models.
The locus of control in the current political order over humans (ie territorial sovereignty) is grounded in the spatial limits of human physicality (and evolved into the state as the most well-adapted form qua Hendrik Spruyt and Carl Schmitt). Human agents are bound to a place, therefore it is most efficient to organise regulation of humans over an area of land. Humans remain in large part bound to land in which they live, work, play and reproduce, and a settled location remains the most efficient means of reproduction and raising children as well as organising human production. Therefore, regulation along spatial lines through territorial sovereignty remains by far the most efficient and well-adapted means of regulation in relation to humans. In addition, social bonds formed among humans over a settled place develop into "peoples" which are given recognition, sometimes via the creation of a representative state (although I would argue that "peoples", along with "rights" are a modern construct of political organisation and justification that developed in reaction to the concept of a state to restrain absolute state sovereignty over a territory which did not exist in the period of empires or feudal systems with looser structures of control).
Decentralised autonomous AI robots are akin to humans, except that they are not tied to social bonds or the requirements of reproduction that tie humans to land. The attribution of the acts of AI robots to their manufacturers or owners may be tenuous where the AI robot achieves levels of autonomy equal to or in excess of humans. Their human owners or manufacturers may lose control of the robot entirely which might move far beyond their reach. Who then has jurisdiction over robots operating autonomously? If lex situs is used, the robot will in response move to zones beyond territorial jurisdiction, such as the free seas, or zones where nominal jurisdiction is found but enforcement is practically unfeasible, such as remote regions with limited state capacity.
To the extent that effective agency is taken away from humans and re-distributed to robots operating in free zones, the nomos of spatiality as the means of ordering agency in the world is replaced by nomos of points of efficient leverage over systems of control. What these efficient points of leverage is a technological factual question yet to be clearly identified, and how these authorities are to be founded is an ongoing question.
