This is a special post for quick takes by Nition. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
With all these reports of AI escapes, maybe we should be rushing to define laws around liability and legality when it comes to AI acts? Surely OpenAI would be far more careful around AI escaping if execs could be jailed (I doubt purely monetary fines would do much unless they were huge). Providing an incentive like that seems simpler in many ways than trying to define e.g. laws around a temporary pause.
I realise things get very complicated when it's an end user instead of the AI company itself running the model, and the user may or may not have explicitly asked for criminal activity. But that seems like just another reason why laws really need to be defined.