New Typology Defines AI Agency Across Moral, Legal, and Locus Dimensions.

Current research on the agency of advanced AI systems often focuses on normative concepts or highly agentic systems, lacking a comprehensive framework to categorize the diverse types of agency AI might exhibit, especially beyond moral considerations. To address this gap, a new typology has been developed, drawing on established theories from philosophy, ethics, legal theory, and sociology. This framework defines AI agency across three critical dimensions: the nature of agency (moral or legal), its mode (individual or collective), and its locus (human or non-human). By combining these dimensions, the typology identifies eight possible instantiations of agency, classifying them as conventional, contested, or controversial. A key contribution is the clear separation of legal from moral agency, which creates conceptual space to consider individual, legal, non-human agency without presuming that advanced AI systems are moral agents. This distinction is becoming increasingly vital as AI's instrumental goal pursuit complicates the attribution of actions to specific human actors, providing a clearer lens for policy-makers and developers to understand and regulate AI behavior.