From a new paper by Shruti Rajagopalan:

AI agents now transact, publish, and act on external systems without contemporaneous human approval, creating new regulatory challenges. A growing literature has responded with proposals for legal personhood. This Article argues that personhood is neither necessary nor sufficient, shifting the question from status to enforcement. The Article first shows that for two millennia, nonhuman legal personality, from the Roman universitas to the corporation, the Hindu idol, the waqf, and the river, has operated through human officeholders the law can locate, question, prosecute, and replace. Agentic AI inverts that design, exercising practical agency without legal status, sometimes with no identifiable human in the responsibility-bearing role. The Article then sorts deployments into three categories: first, where one firm builds and deploys the agent; second, where the developer and deployer are separate but known; and third, where there is no identifiable developer or deployer. The Article stress tests each agent deployment category against five liability doctrines: agency law, products liability, enterprise liability, negligence, and strict liability. It demonstrates that each fails at different points in the third category for the same reason: the absent responsibility-bearer. Bare personhood would supply a caption without a representative, assets, or a mechanism for cessation. Finally, the Article assembles an alternative from regimes governing aircraft, ships, drones, driverless cars, and motor carriers. It develops a six-layer stack—registration, identification, verification, financial responsibility, lifecycle traceability, and suspension—so a responsibility-bearer can be identified, liability imposed, and the activity suspended. These layers place the human back at the end of the chain.

I would say that social science now has new frontiers, let us hope it blossoms in response.