General Autonomy: Enabling Organization from Cells to AI Alignment

Life is organized across multiple scales of autonomy, from cells to societies, and remains robust because the self-maintaining organizations at each scale stay largely compatible with one another. We develop a General Autonomy framework in which one relation accounts for autonomy, agency, and alignment alike: enabling relations, the contributions processes make to one another's persistence. Self-maintenance closed within a unit grounds autonomy; enabling paths that return through the environment ground agency; compatibility among self-maintaining units grounds alignment. These are substrate-independent principles of general autonomy, from cells to AI alignment, with life as the paradigm case. We give a minimal formalization in which contributions are defined interventionally, through their effect on a system's viability. Two quantities grade autonomy: the strength of a unit's self-maintenance and its robustness, the depth given by how long an interruption of its own self-production it can recover from, which separates shallow attractors such as vortices and flames from deep organizations such as cells. Alignment then follows from autonomy rather than opposing it: under three structural conditions—reciprocal support from the whole, the unit's dependence on it, and support conditional on its contribution—a self-maintaining unit constructs alignment with the whole it belongs to, because contributing is its own least-cost route to persisting; each condition fails characteristically. Applied to AI, this recasts alignment as an organizational problem: external control remains useful but becomes structurally insufficient as agents grow more autonomous. We propose enabling-based alignment, shifting the target of design to the organization that sustains an agent: its own self-maintenance, embedded in networks of mutual dependence with humans and institutions.