Relational Sovereignty: Why Autonomous Systems Need Each Other to Stay Distinct

What makes a relationship between intelligent systems a good one? Philippe argues it’s a balance between being predictable enough to trust and surprising enough to matter. A key ingredient of that is agent sovereignty (freedom from interference with an agent’s data breaks its self-model), distinctiveness (a system identical to its peers collapses into them), and autonomy (a sovereign system distinguishes itself best when it acts for itself). It extends the relational framework from his Cohere Labs talk, Could phenomenal reports lead to better multiagent collaboration?, into the territory of sovereign, adversarially robust agent networks.

Links