Answers are grounded in this paper and cite their sources.
Agentic AI systems now act, rather than merely advise, within social and institutional settings. The persistent gap lies less in principles than in deciding which obligations a deployment should attract as the relationships around it change. Risk-based regulation classifies by declared intended purpose and does not register the dependence that forms around a system once it is in use. We propose the Onto-Relational-Sophic (ORS) framework, whose core contribution is to treat that dependence, rated as social embedding, as a trigger for reclassification. A system-in-deployment is rated on autonomy, social embedding, and moral relevance; each rating triggers a module of obligations on providers, deployers, and institutions; nine explicit rules govern assignment, conflicts, and reclassification, including the rating of multi-agent ensembles as wholes and the prospective rating of systems designed to induce dependence; and categories serve as summary labels. Two supporting pillars, grounded in Cyberism, structure the framework: a Cyber-Physical-Social-Thinking ontology that organises the evidence for each rating, and Cybersophy, an axiology that applies a reversibility priority when principles conflict. Three scenarios and a structured comparison with existing EU law show where ORS converges with current obligations and where it adds to them, most clearly when people come to depend on one certified system in different ways. No legal personality is conferred on any system. The rating scales and rules are proposed rather than validated: the social-embedding instrument has not been tested for inter-rater reliability, and the worked classifications rest on the authors’ own ratings. A reliability study should precede regulatory use.
No takes yet. Share an insight, caveat, or question.
Ning et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: