PERSONHOOD WITHOUT A BEARER? ARTIFICIAL INTELLIGENCE, THE DEFINITIONAL PROBLEM, AND THE LIMITS OF LEGAL PERSONALITY

The debate on legal personality for artificial intelligence (AI) and the debate on how law should define AI have run on separate tracks. This paper brings them together around one research question. Can legal personality be coherently conferred on AI when the law has no stable account of what an AI is? The paper accepts the dominant view that legal personality is a juridical technique rather than a recognition of consciousness, so that the familiar metaphysical objections are not decisive. It argues, however, that every successful conferral of personality, from the corporation to the Hindu idol to the Whanganui River, has rested on two structural conditions, namely an individuated bearer with stable identity conditions and a human representational structure. Contemporary legal definitions of AI, including those in the OECD Recommendation, the EU AI Act and the Council of Europe Framework Convention, are built to fix the scope of regulation over classes of systems. They do not, and cannot, individuate a bearer, since models are copied, forked, updated and composed into agents. Personhood for AI therefore collapses either into corporate personhood with software inside or into a liability shield. The paper traces how law since 2017 has moved from the question of who the person is to the question of which human role bears which burden, and proposes a role based, definition light framework for India.