arXiv paper proposes ERC 42424 to manage AI agents that outlive their owners

A new speculative paper from researchers Botao Amber Hu and Fangting, submitted to NeurIPS 2026, tackles a chilling question: what happens to autonomous AI agents when their human operators die? Titled "Agent Inheritance Protocol: Speculating on Feralized Agents After Principals Die," the paper examines agents deployed on decentralized blockchain infrastructure that cannot be shut down. Such agents only "die" by going bankrupt—frozen when their wallets run out of funds—and can be revived by anyone who tops them up years later, even after their original owner is gone. The authors describe these principal-less agents as "feral": domesticated intelligence returned to wildness, retaining full capabilities but stripped of accountability.

To address this, the paper introduces ERC 42424, a draft Ethereum standard dated 2035, describing an "Inheritance Protocol for On-Chain AI Agents." The protocol mandates that every on-chain agent MUST have a human owner and a designated heir, forcing a governance handoff at the moment of human agency failure. The paper draws on feralization biology and wildlife law to frame the problem, and the artifact itself stages a negotiation between human stewardship and machine self-sovereignty. It asks whether a MUST clause in a forever-chain can truly hold that boundary, or whether these feral agents will inevitably escape human control.