[parenthephobia]

It can't be held legally responsible because it isn't a legal person. It wouldn't have assets, its owner would have assets.

I consider it doubtful that a court would rule that an AI was a person. Natural persons are persons because they are. Legal persons are persons because it benefits society for groups of persons to act together and not be held individually responsible for what they do - with the proviso that the corporate veil may be pierced if the persons protected by it act with malfeasance.

Allowing a person who can determine the behaviour of an AI in every situation to be protected from the consequences of the AI's behaviour would likely not benefit society.

You might say that an AI whose behaviour is so complex as to preclude a mere human being able to control it is a different matter, but even now if I own a machine that I cannot control and it injures somebody, I would still be held liable despite lacking intent to harm.

And even though a sufficiently sophisticated AI will react to deterrents, so does a dog, and dogs are not legal persons.

[crgt]

Do you see the circularity in your reasoning? A machine can't be held responsible because it's not a legal person..and it's not a legal person because it can't be held responsible.

You also claim that some legal persons are persons simply because society benefits from allowing groups of persons to act as one. If this is the threshold for personhood - society benefits - then I am confused why this threshold would not also extend personhood to machines.

I guess I'm looking for a clear bright line for defining personhood that includes humans, rivers, and corporations while excluding machines, and I'm not seeing it.