> If we don't establish strict liability now, we're in for an era of stochastic crimes that go unpunished for anyone who is not rich or a large corporation. I

> If we don't establish strict liability now, we're in for an era of stochastic crimes that go unpunished for anyone who is not rich or a large corporation. I very much agree with this - making AI companies explicitly responsible if their internal AI causes hacks etc could do a lot to improve their safety considerations. But I wonder what the liability should be when it's a third party using the AI and that AI hacks, intentionally or not. If a users tells ChatGPT to hack something and it succeeds, is the user the person responsible because they told the AI to hack, in the same way Victorinox is not responsible if you stab someone with one of their knives? Or is OpenAI to some extent responsible as well since they made a powerful tool without sufficiently strict safeguards? What about if the user was trying to do something legal and the AI made the decision to hack by itself?