22 June 2026 · LinkedIn
In 1602 we built a person that cannot think, feel, or die, purely so that someone could be sued. Some now argue we should do the same for AI.
That first one was the Dutch East India Company. Legal personhood let it own property, sign contracts and be held liable as a body, not as the people inside it. It was plumbing for responsibility: when something went wrong, a court had someone to reach.
That is the precedent some are now reaching for: that as AI agents take on roles once held by people, they should hold legal standing of their own, even stakeholder status inside the firms they work for. The argument keeps snagging on whether a model can really think or feel. That was never what personhood was about. The real question is liability and standing, and who is left holding them.
Today an AI is a tool, and the company that deploys it owns the consequences, which keeps a named human on the hook. Grant it personhood and that flips: the AI becomes the thing holding the liability.
But responsibility needs something that persists, and here the ladder runs the wrong way. A person carries one self through time, imperfect but continuous. A company carries less, its directors change, but the body endures. A model carries none. Copy it a thousand times, and even one copy keeps no thread between sessions, nothing that remembers the deed or could answer for it.
So the liability lands on a void, and the company that deployed it walks away clean. We built the legal person so the buck always stopped somewhere. This would be the first one built so it stops nowhere.
Join the conversation on LinkedIn →
← All writing