Heather Alexander (Tilburg University), Jonathan Simon (New York University — Department of Philosophy), and Frédéric Pinard (University of Montreal) have posted How Should the Law Treat Future AI Systems? Fictional Legal Personhood versus Legal Identity on SSRN. Here is the abstract:
Today, artificial intelligence (AI) systems are more agentic, more autonomous and more unpredictable, than ever. A major, and underexplored, risk of agentic AI is the threat it poses to the coherence of laws and legal system. One solution is to reclassify some future AI systems as persons rather than as objects under the law. We argue that fictional legal personhood is not fit for purpose, explain why legal identity is the better option, and discuss why hybrid approaches won’t work.
Highly recommended. Alexander, Simon, and Pinard engage the trustee argument from my 1992 paper, Legal Personhood for Artificial Intelligences, 70 N.C. L. Rev. 1231 (1992). They argue that legal persons require clear identification, registration, and continuous identity, but that AI systems can be copied, versioned, fine-tuned, merged, or deleted — raising the question of which instance or version bears the legal personhood. Under the existing legal framework a trustee is a legal role occupied by an artificial (corporation) or natural (human) person who holds legal title to the trust assets, subject to fiduciary obligations imposed by the trust instrument. Succession is handled by legal rules governing the transfer of trusteeship: when a trustee is replaced, legal title transfers to the successor. Applied to AI: the copying and versioning problem is answerable because legal title and fiduciary obligations attach to the designated, registered trustee — exactly one legal person holds title at any time. The others, however similar in their underlying model weights, are simply not the trustee. Retraining raises the ship-of-Theseus problem, but corporate law handles analogous questions through formal institutional continuity rather than metaphysical identity: a corporation that substantially restructures remains the same legal person if continuity of registration is maintained. The same principle applies to an AI trustee — what matters is the formal designation, not questions about the technical identity of the underlying system. I have to think more about the important issues that Alexander, Simon, and Pinard raise, but at this point I am inclined to think that a legal identity framework doesn’t provide the functional advantages of some form of legal personhood for specific legal roles — assuming that an AI has the right functional capacities.
To receive a daily summary of posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.