How Should the Law Treat Future AI Systems? Fictional Legal Personhood versus Legal Identity

Current AI systems should remain classified as objects under law, not persons. The authors find that an "object approach" remains adequate for existing AI as of 2025, suggesting that legal personhood frameworks are premature for today's technology.

The law may eventually need to distinguish between two types of AI legal status: "fictional personhood" (similar to corporations, with limited and revocable rights) versus "non-fictional personhood" (with fundamental, non-derogable rights like those afforded to humans). This distinction matters because it determines what rights and protections an AI system could claim.

Granting AI systems legal personhood would have far-reaching implications across multiple legal domains, including liability frameworks, intellectual property, family law, civil rights, and citizenship—suggesting that any such change would require comprehensive legal reform, not piecemeal adjustments.

The authors tentatively conclude that "non-fictional personhood" may ultimately provide better long-term legal coherence than maintaining an object classification for sufficiently advanced AI systems, though they acknowledge this remains a complex philosophical and practical question requiring further deliberation.

Legal personhood decisions should be tied to specific technical capabilities and characteristics ("suitably advanced, individuated AI systems"), not granted universally to all AI, implying that future policy should establish clear criteria for when and how AI systems might qualify for different legal statuses.