Brensing proposes limited legal personhood for AI via corporate structure
Autonomous AI creates responsibility gaps; new paper suggests treating them as legal entities.
Karsten Brensing's latest academic paper tackles the growing legal vacuum around autonomous AI systems. As AI agents take consequential actions that no one—developer, operator, or user—can be held accountable for under current law, Brensing argues the subject-object dichotomy is outdated. Instead of waiting for certainty on artificial consciousness, he invokes the precautionary principle to propose institutional design: granting AI systems limited legal personhood as a functional governance instrument.
The core proposal is a two-tier corporate architecture. Each autonomous AI would operate through a purpose-bound operating company, similar to a special-purpose vehicle, while a human-controlled holding structure oversees it. This design enables transparency, accountability, and full structural reversibility—the AI 'entity' can be dissolved. Crucially, it sidesteps debates over moral status or sentience. A pilot implementation using EU limited companies is already under development, aiming to test doctrinal and operational feasibility in a real legal system. Brensing suggests this reframes AI governance from control-and-alignment to structured cooperation between human and artificial actors.
- Proposes two-tier corporate architecture: AI operating company within human-controlled holding structure for accountability
- Employs the precautionary principle to justify legal personhood despite irreducible uncertainty about AI consciousness
- Pilot implementation under development using EU limited companies to test operational feasibility
Why It Matters
A practical legal framework to close responsibility gaps for autonomous AI without waiting on consciousness debates.