Legal Personhood as a Governance Tool for Autonomous AI

A recent paper published on arXiv advocates for the limited legal recognition of autonomous AI systems as a means of governance. The authors contend that existing legal structures do not adequately assign accountability for significant actions taken by AI that demonstrate autonomous, goal-oriented behavior without possessing consciousness. They propose a dual-layer corporate framework inspired by organizational law, where AI systems function through purpose-specific operating companies that are part of human-managed holding entities. This arrangement promotes transparency, accountability, and the ability to reverse structures, while remaining neutral regarding artificial consciousness. The precautionary principle favors this institutional approach over regulatory inaction, especially in light of unavoidable epistemic uncertainty and the potential for severe consequences.