Darwinian Emergence of Minimal Functional Agency in AI Systems

We present a theoretical framework establishing that minimal functional agency emerges inevitably in artificial intelligence systems satisfying basic structural conditions. Drawing on Darwinian principles — random variation coupled with selection — we demonstrate how current AI architectures exhibiting stochastic exploration, reinforcement mechanisms, persistent state, and open-ended interaction develop stable, goal-directed, self-modifying behaviors that satisfy established philosophical criteria for agency. We formalize minimal functional agency (MFA) through four observable conditions: persistent objectives, policy continuity, counterfactual sensitivity, and self-generated subgoals. MFA requires neither consciousness nor moral status, being defined purely at the level of functional organization.